Who is internationally diversified? - Workspace

Who is internationally diversified? Evidence from the 401(k) plans of 296 firms Geert Bekaert*, Kenton Hoyem+ , Wei-Yin Hu+ , Enrichetta Ravina* April 2016 Abstract Drawing on a novel database of the 401(k) plans of 296 firms, we examine the international equity allocations of 3.8 million individuals over the period 2005-2011. We find enormous cross-individual variation, ranging from zero to more than 75%, and strong cohort effects, with younger cohorts investing more internationally than older ones, and each cohort investing more internationally over time. Access to financial advice, lower fees, and more international fund choices are associated with higher international allocations, suggesting a role for plan design and policy. Education, financial literacy, and the fraction of foreign-born population in the ZIP code also have positive effects on international diversification, consistent with explanations based on familiarity bias and information barriers.

*Columbia Business School +

Financial Engines, Inc.

JEL Codes: G11, G15, G23 Keywords: International Diversification; Micro Data; 401(k) Portfolios We thank Dan Amiram, James Choi, Jonathan Reuters, Frank Warnock, an anonymous referee and seminar participants at the American Economic Association Annual Meeting, the Annual Conference in International Finance at Imperial College, the Federal Reserve Board, the University of Texas at Dallas, the 27th Australasian Finance and Banking Conference, the Norwegian Financial Research Conference, and the 2014 Retirement Research Consortium, the EIEF, the World Bank and the 2015 WFA Annual Conference for their helpful comments. Nicolas Crouzet, Katrina Evtimova, Jia Guo, Yihua Lin and especially Andrea Kiguel provided excellent research assistance. This article reflects the findings and opinions of the researchers and not necessarily those of Financial Engines. Financial support from the Chazen Global Research Fund, the Sandell Grant from the Boston College Center for Retirement Research and the Social Security Administration is gratefully acknowledged. Corresponding author: Enrichetta Ravina, Columbia Business School, 3022 Broadway, Uris 822, New York, NY, 10024, tel. (212)854-1031, fax (212)6628474, email: [email protected].

Electronic copy available at: http://ssrn.com/abstract=2417976

Introduction The proportion of domestic stocks in most investors’ equity portfolios well exceeds their country’s relative market capitalization in the world, which causes investors to forego substantial diversification benefits.

This home-bias phenomenon remains one of

international finance’s major puzzles. An increasing number of studies have investigated the determinants of home bias from both rational and behavioral perspectives (see Sercu and Vanpee, 2012, for a survey). The country-level international under-diversification documented in the literature masks much individual heterogeneity. Table 1 shows some statistics for international equity allocation (as a percentage of the total equity allocation) of 3.8 million U.S. individual 401(k) accounts across 296 different firms over the period 2005-2011 period.1 We stratify the data into older people (born in 1960 or earlier) and younger people (born in 1980 or later), and contrast average international allocations for either the 5 most internationally diversified firms relative to the 5 least diversified firms, or the most diversified state (Iowa) relative to the least diversified state (Nevada).2 Irrespective of the 1 A 401(k) plan is a defined-contribution retirement savings plan offered by many U.S. firms to their employees (“401(k)” refers to the subsection of the Internal Revenue Code that defines the plans). Employee contributions are made as deductions from their paychecks and placed in an individual account for each employee within the plan. The firm typically provides a range of investment options from which each employee can choose. Funds in these accounts receive a variety of different preferential tax treatments, and may also receive matching contributions from the firm. 2

For each firm (state) we calculate the average international equity allocation (as a percentage of the total

equity allocation) in the 401(k) accounts of the employees working at the firm (residing in the state) and rank the firms (states) from highest to lowest.



2 Electronic copy available at: http://ssrn.com/abstract=2417976

salary group (we considered three groups), people in Iowa have about 5-10% higher international allocations than people in Nevada; the difference for diversified vs. nondiversified firms is even larger, at 20-30%. Moreover, older people are consistently less internationally diversified than younger people. Our analysis of this cross-individual dispersion provides a unique perspective relative to the related international finance literature, which has primarily used cross-country data on asset holdings to uncover various determinants behind home bias.

Research has

documented both host and destination (target) country factors behind these biases, but the focus has mostly been on destination country factors, such as corporate governance issues, stock market development, and investment restrictions. 3 To identify these destination country factors, studies then focus on the related problem of foreign investment bias, examining to what degree home-biased countries underinvest in various countries. Particularly popular are explanations based on information barriers (Ahearne, Griever and Warnock, 2004; Brennan and Cao, 1997; Van Nieuwerburgh and Veldkamp, 2009) and familiarity biases (Portes and Rey, 2005). For comparison, Bekaert, Siegel and Wang (2013) document the cross-country dispersion in home bias relative to a CAPM (relative market capitalization) benchmark for 35 countries, normalized to be between 0 (no home bias) and 1 (all equity holdings in 3

The determinants proposed by those studies include transaction costs (Glassman and Riddick, 2001), real

exchange rate risks (Fidora, Fratzscher and Thimann, 2007), information barriers (Ahearne, Griever and Warnock, 2004), corporate governance issues (Dahlquist, Pinkowitz, Stulz and Williamson, 2003; Kho, Stulz and Warnock, 2009), stock market development (Chan, Covrig and Ng, 2005), the need to hedge local consumption streams (Aviat and Coeurdacier, 2007), investment restrictions (Bekaert, Siegel, Wang, 2013), and lack of familiarity (Portes and Rey, 2005), to name a few.



3 Electronic copy available at: http://ssrn.com/abstract=2417976

domestic stocks). The least home-biased developed country is the Netherlands, with a home bias for the period 2001-2009 of only 34.7%, while Spain, the most home-biased, has a home bias of 87.5%. It is straightforward to convert the numbers of Table 1 into relative home-bias numbers (we divide by the fraction of world market capitalization accounted for by non-US markets, which is 64.4%, and subtract that ratio from 1). For a “1960” cohort person with median salary at a poorly diversified firm, normalized home bias is 92.45%; in contrast, it is only 43.63% for a “1980” cohort person at a relatively well- diversified firm, indicating that the cross-individual dispersion of home bias within the U.S is of the same order of magnitude as the cross-country dispersion. Understanding this cross-individual dispersion may have profound implications for the international diversification literature. First, pure destination country factors, such as various investment restrictions in different countries or corporate governance problems, which are inherently difficult to measure, cannot explain the cross-individual variation in international diversification for U.S. individuals. Second, the cross-individual dispersion suggests that individual heterogeneity in preferences or background risk may play a large role in driving international underdiversification and be more important than the “cost” of international investing or international risk factors, such as transaction costs and real exchange rate risk.4 Personal characteristics such as age, salary, and wealth may play a role. Familiarity bias (Huberman, 2001) or informational asymmetry between local and nonlocal investors (Coval and Moskowitz, 1999) also have implications for the incidence of “international” home bias for individuals in different locations within the U.S. (e.g., 4

There may, of course, be variation in the quantity, quality, and diversity of the foreign investment options

in different 401(k) plans, and we explicitly examine their effects in Section V.



4

based on the number of foreign-born people in a region), as may working for different firms (international vs. domestic firms).5 Finally, cross-country studies miss a set of potentially important determinants of home bias that may be policy relevant, such as education level and the quality of the investment options available to the individual. Each individual in our sample can be described by personal characteristics, the area where she lives (captured by ZIP code), and the firm she works for. We therefore proceed in three steps. We first analyze the importance of personal characteristics such as age, cohort, salary, and wealth indicators, as well as access to financial advice. From these regressions, we identify ZIP code and firm fixed effects and analyze these separately. Fortunately, several of the firms in our sample are large firms with multiple branches in different locations, in some cases spread out over the whole country. This enables us to meaningfully differentiate location from firm effects. One key fact emerging from the data is that there is an upward trend in the extent of international diversification. We show that part of this—but only a small part—is the potentially rational response to the slowly decreasing importance of the U.S. market in the world equity markets. We also find negative age and positive cohort effects. As is well known, time, age, and cohort effects cannot be separately identified (see Ameriks and Zeldes, 2004). We argue that the most plausible characterization of the data is a strongly positive cohort effect coupled with a pure time effect, as opposed to assuming that investors decrease their international allocations as they age and that this decrease is counteracted by an overall trend toward more diversification. The cohort effect is 5

See Brown et al. (2015) for an application of the information advantage story to the local tilt of the equity

portfolios of state pension plans.



5

partially responsible for the trend toward more international diversification over time. In addition, each cohort invests more internationally each year, which delivers the strong upward trend in international diversification (see Fig. 5). Trend and cohort effects are consistent with the ongoing globalization process, which makes people more comfortable with foreign investments over time. As alternative explanations for the cohort effects, we also analyze the role of past return experiences by investigating various return-sensitive variables, such as the Malmendier and Nagel (2011) stock market experience effect, adapted to the international investment setting by replacing the stock market return with the foreign minus the U.S. equity return, or the foreign return alone (return chasing), based on the idea that people who have experienced higher relative or absolute international returns may be more likely to invest internationally. Similarly, we explore the effect of investor inertia by determining whether the cumulative returns of foreign vs. U.S. equities experienced by the individual between different dates explain the degree of international allocations. Finally, we consider the role of periods of stress in the markets, which might induce investors to become more home-biased. The analysis is reported in Section II, and we conclude that a simple cohort effect best explains the data. Among individuals’ personal characteristics, we find that higher salaries and higher house values (measured by the median house value in the ZIP code in which the individual resides) are associated with higher international allocations, while higher account balances are associated with lower international allocations. Only the salary effect, however, is economically meaningful.



6

A critical variable correlated with the degree of international diversification is access to financial advice. About 11% of the individuals in our sample signed an investor service agreement to receive online advice from Financial Engines, and approximately one third of those access the advice website regularly. Individuals who signed up for advice and have accessed the website recently have international allocations that are 5.3% higher than those who never signed up, and 2.2% higher than those who signed up but have not accessed the website recently. Interactions of access to advice with demographics indicate that advice is more strongly associated with higher international allocations for older cohorts—the cohorts that, based on our results, would otherwise invest less in international stocks. Additional demographic characteristics are captured in the ZIP code and the firm fixed effect analysis. In studying ZIP code effects, we find that higher education levels are associated with significantly higher international equity allocations, both statistically and economically, and that the same is true for financial literacy. For example, shifting the proportion of people with a high school diploma in the ZIP code from the 25th to the 75th percentile of the distribution in the sample generates a 0.48% increase in international allocations. Making the same shift for bachelor’s degrees adds a further 0.70%, and an advanced degree 0.44%. Cumulatively, the effect is about 1.62%. Similarly, going from poor to high financial literacy amounts to a 0.47%-0.88% increase in international diversification over and above the effect of education. We also find evidence that is potentially consistent with the familiarity hypothesis. ZIP codes with a higher percentage of individuals born in foreign countries have higher international allocations, even controlling for the average (median) house value per ZIP code and for state GDP growth



7

and levels, although the effect is more modest. Also, consistent with the familiarity or information hypotheses, we find that more export-oriented states feature higher international allocations. A firm‘s culture or the firm’s activities may make employees more familiar and comfortable with investing internationally, but may also change the risk and characteristics of their human capital. For this reason, we attempt to measure how “international” a firm is, either directly or indirectly. We control for whether the company is public, whether the parent is incorporated locally or abroad, and for the presence of foreign subsidiaries. We also control for firm size, leverage, profitability, and sales and investment over assets. Firm fixed effects reveal that employees of profitable firms invest less, and employees in private firms and in firms with foreign subsidiaries invest more in international equity. Another important dimension through which a firm affects the international allocations of its workers is training sessions, social interactions, the investment options available, their quality and fees, and their evolution over time. To control for these features, we rerun our baseline specification, adding fixed effects based on the quarter the individual joined the firm interacted with the firm’s identity and further interacted with the quarter-year of observation. Results indicate that most effects remain robust across specifications, but the cohort effect is halved. Further analysis controls directly for the fraction of international funds among the equity funds offered by the plan, expense ratios relative to the domestic equity funds in the plan, relative turnover, fund age, expense ratios of the funds in the plan compared to all the funds in their same category, difference between historical



8

alphas of international vs. domestic equity funds in the plan, and total plan size.6 Results demonstrate a significant association between international diversification and plan features. For example, shifting the fraction of international funds by the interquartile range of this variable, from 16.7% to 25.6%, is associated with an increase in international allocations between 1.57% and 1.81%, depending on the specification. In addition, lowering expense ratios from the 75th percentile to the 25th percentile of their distribution is associated with about a 0.75% higher international allocation. To the extent that plan features are determined by the employers and are not exclusively the result of employees’ demands, the findings above indicate that improving the quality of the international investment options offered by 401(k) plans, in terms of number of funds available and their fees, generates more investment in international equity. Finally, it is possible that some of our investors invest internationally by holding U.S. stocks that have more exposure to international factors, e.g., multinationals (Cai and Warnock, 2012). This is unlikely to be the case in our sample, however, as the correlation between the international equity allocation and the allocation to large U.S. equities—the category to which multinationals likely belong—is actually positive and increasing over time. Moreover, extensive research suggests that multinationals alone are not sufficient to span the international diversification benefits of investing in foreign companies (see Lewis, 1999; Rowland and Tesar, 2004). Hitherto, the large majority of home-bias studies are based on aggregate statistics, while individualized perspectives on home bias are largely limited to studies of Swedish households by Calvet et al. (2007), Karlsson and Norden (2007), and Norden (2010). 6



We obtained these detailed data for all the plans offered by the firms in our sample for 2012 only.

9

Calvet et al. (2007) do not specifically focus on international diversification, but they note that Swedish households are relatively well diversified internationally because popular Swedish mutual funds have a high international allocation. Karlsson and Norden use a sample of 9,415 Swedish individuals for the year 2000 to study the likelihood of home bias, and find that wealth affects home bias negatively and age positively. Norden (2010) shows that underdiversified investors are worse off than those who are well diversified internationally, but the latter’s advantage is diminished by a proclivity to excessively churn their portfolio. Graham, Harvey, and Huang (2009) use a UBS survey of 1,000 investors to demonstrate that investors who feel competent trade more often and have more internationally diversified portfolios. The remainder of the article is organized as follows. Section I describes the data and some summary statistics. Section II investigates the effect of personal characteristics and time effects on international diversification, while Section III focuses on geography and Section IV on firm effects. Section V investigates the effect of plan quality and menu design on international diversification, and Section VI reports a number of robustness checks aimed primarily at confirming that the account variation we rely on mostly reflects portfolio variation at the individual level. Section VII concludes.

I.

International Diversification at the Individual Level

Data description For this study, we use a large proprietary dataset, comprising approximately 3.8 million individuals, made available by Financial Engines, which is the largest independent



10

registered investment advisor in the U.S. and provides advice and investment management to 401(k) plan participants. The dataset includes record-keeper information on demographic characteristics, balances, salary, 401(k) contributions, household ZIP codes, and the “style” of the asset allocation (see Sharpe, 1992), divided into 5 aggregated asset classes and company stock. The underlying style analysis applied to the funds in each plan initially uses 15 asset classes spanning domestic and foreign equities and fixed income investments. Style analysis finds asset class weights such that the residual return (the difference between the actual fund return and the style return) has minimal variance, with the weights adding up to 1 and constrained to be non-negative. One of the 5 aggregated asset classes is “International Stocks,” and its underlying styleanalysis model uses indices on European, Pacific, and Emerging stock markets. Data are drawn every quarter, with a given individual being sampled approximately every 6 months. For a limited number of companies, the data sample starts in 2005, but the sample becomes much more complete during the second half of 2006 and runs to the end of 2011. In addition, we have detailed information on various features of the 401(k) plans, such as plan size from IRS Form 5500, and on the plans’ investment options for a more recent subperiod. The final dataset combines proprietary data from Financial Engines on asset allocations, contributions, and demographics; information on 401(k) plans’ menus and features; financial information on the companies from CRSP, Compustat and CapitalIQ; and census and other sources of socioeconomic data matched through household ZIP codes. An important consideration is whether individuals in our sample receive financial advice, particularly because Financial Engines provides financial advice and asset-management



11

services to the firms we have data on. Two types of advice are relevant for our purposes. First, under the “managed account” model, Financial Engines simply manages the portfolio on behalf of the client and charges a fee on the assets under management. We exclude these individuals, as the international allocation in such accounts is high, but is set by Financial Engines. Second, Financial Engines also provides online advice, which the client must implement himself. We have information on both when the individual joined the online advice program and when he last logged on, and control for this form of advice in our regressions. Finally, our data include separate information on the allocation to target date funds (TDFs), and we control for this in our analysis as well. Our sample contains 296 firms. In Appendix Tables 1 and 2, we report some characteristics of the firms and workers in our sample and compare them to firms in Compustat and the S&P 500 Index, and to the population of full-time U.S. workers as reported in the Current Population Survey (CPS). In terms of size, whether we look at assets, sales, or number of employees, the firms in our sample are substantially larger than the Compustat firms. Average net income and capital expenditures in our firms also exceed that of Compustat firms. For example, the median number of employees is about 4,600 in our sample, while it is only 475 in the Compustat sample; the average number of employees per firm is more than 18,000 in our sample and about 8,000 in Compustat. The presence of such large companies means that the employees of one firm may be geographically dispersed across the country. Our firms have higher ROAs, but their leverage ratios are similar to those of the companies in Compustat. Average annual returns are higher in our sample and the dispersion is rather wide, due to the financial crisis. Compared to firms in the S&P 500, the firms in our sample are smaller, with



12

slightly smaller asset size but far fewer employees.

Our companies are mostly

established companies, with the median age being 65 years and the interquartile range 28103 years. Finally, in Panel D we contrast the characteristics of the private and public firms in our sample. The public firms in our sample are larger in terms of assets, sales, and number of employees. However, the private firms are not small upstart companies. Their median age is 42 years, and the median number of employees is about 13,000. The average plan size is large—roughly $500 million on average—but because there are many small plans as well, the median size is only about $300 million. In Appendix Table 2, we compare worker characteristics in our sample with those of fulltime workers in the overall population. The workers in our sample tend to have higher salaries, with the average and median salaries being around $15,000 to $20,000 higher than in the population at large. The average tenure is also about 5 years longer. Finally, the workers in our sample are on average about 4 years older. Salary shows a smooth concave pattern with respect to age: almost linearly increasing at first, then flattening out around the 51-55 age group, with salaries starting to decrease for people aged over 60. We also report account values for our sample, which have a very skewed distribution with the mean at $70,000 (higher than the average annual salary) and a median value of $25,786 (which is actually lower than the median annual salary). Account values may reflect a mixture of tenure, past salaries, and contribution rates. Contribution rates vary between 0% and 17% of salary, and are on average equal to 6%.

Measuring International Diversification



13

We start with some simple notation. Let wintt,i be the allocation to international equities of individual i at time t and weqt,i her allocation to all equities (domestic and foreign equities). Our main variable of interest is the extent of international equity diversification, idivt,i = wintt,i/ weqt,i. The international home-bias literature has used a wide range of measures, including international holdings over GDP (Aviat and Coeurdacier, 2007), or portfolio flows scaled by market capitalizations (Portes and Rey, 2005). Our focus, however, is on portfolio choice, so that the international equity allocation is the natural variable to focus on. A number of articles (for example, Ahearne et al., 2004) have used relative weights, controlling for what the allocation would have to be under, typically, a simple World CAPM benchmark. Such relative weights also partially control for international vs. local valuation changes. We use such a CAPM benchmark weight in our empirical analysis, but focus on the actual extent of international equity diversification as our main variable of interest. Bekaert et al. (2013) study several biases that plague standard measures, including size biases that arise from the fact that countries with a relatively large market capitalization are mechanically less likely to be severely home biased on a relative basis than countries with a small market capitalization. However, because we focus on allocations from the citizens of one country, we need not worry about such biases. We would also like to characterize the international allocation to bonds but do not have the data, as the aggregated bond asset allocation reported in our dataset does not distinguish domestic from international bonds. This also makes it natural to scale by equity holdings and not by total holdings.



14

Our focus on equity diversification has two additional advantages. First, by focusing on international allocation among stock market participants, we avoid confounding international nondiversification with stock market nonparticipation. Second, the focus on equity allocation potentially circumvents issues raised by optimal asset location. A high bond allocation and low equity allocation may reflect optimal asset location, given that the effective tax rate on bonds is mostly higher than on equities.

Under certain

assumptions, the relative equity allocation should be constant across different accounts, even across taxable and tax-deferred accounts (Huang, 2008), and therefore the idiv variable can be meaningfully examined even in accounts with relatively low equity allocations. Nevertheless, in the robustness section of the paper we replicate the baseline results, excluding individuals with high bond allocations, and find that the results continue to hold. Another tax issue is that some foreign countries levy withholding taxes on dividends and interest income, eroding the advantage of holding international assets in the nontaxable account. We also consider a robustness check that eliminates investors who might engage in such an asset-location strategy by focusing on accounts with high bond allocations for people with relatively high salaries and account balances. Again, our main results remain robust. International Diversification across the U.S. Insert Fig. 1 here: Cross-individual variation in International Diversification In Panel A of Fig. 1, we present a histogram of the international allocations over all of our observations. The average allocation is 17.7%, and 37% of our observations lie between 10% and 25%. In addition, 17% of the allocations are exact zeroes, while 3% of our observations reflect allocations to international equity of more than 50%.

15

The reason the average allocation of 17.7% is usually viewed as “under”-diversification, is that foreign equity markets during our sample period represent, on average, 64.4% of world market capitalization (computed using MSCI data; the MSCI index covers approximately 85% of the free float-adjusted market capitalization in each country). We denote the relative importance of foreign equity markets in world markets as idivt,bm. Note that this benchmark is only optimal under the strict assumptions of the CAPM, but we use it here as a reference point for our analysis. In panel B of Fig. 1, we show the histogram for relidivt,i = idivt,i / idivt,bm. When relidiv is larger than 1, the individual is overdiversified according to the World CAPM benchmark; if it is 1, the individual invests according to existing relative market capitalizations, while 0 represents full home bias. The statistic is bounded from above by 100 divided by the fraction of the world market capitalization represented by foreign equity markets. This bound is 156% when evaluated at the average value of the foreign equity market fraction. Looking at Fig. 1 (Panel B), we see that only slightly more than 2% of the observations are higher than 90%, representing almost full or over-international diversification. Slightly more than 47% of the observations show relative diversification less than 25%. These data are consistent with aggregate data on international diversification. We computed a proxy for the proportion of mutual fund holdings in international equity funds to total equity funds from various editions of the ICI Factbook. The estimates use assets under management in “world” equity open-end funds divided by the total of the world and domestic equity categories from Table 3 of the Factbook. This fraction increases from 23.3% in 2006 to 26.0% in 2011. These numbers likely slightly overestimate the international diversification proportion, as the “world” category also



16

includes global funds, which can invest both internationally and in the US.

Overall,

these numbers appear consistent with ours.7 Insert Fig. 2: Geography and International Diversification Fig. 2 shows the international diversification averages for each state. Aggregating at the state level compresses the distribution considerably, but we still clearly see a spread between relatively well-diversified states (Utah, Iowa, Hawaii) with idiv’s of more than 20%, and poorly diversified states (Alabama, West Virginia, and Nebraska) with idiv’s close to 15%. Insert Fig. 3: Firms and International Diversification In Fig. 3, we show the histogram after averaging idiv and relidiv within firms. One possibility is that the quality and diversity of a firm’s 401(k) plan options is the main driver of the observed cross-individual variation in international allocations. For example, Elton, Gruber, and Blake (2004) study over 400 plans and find them “inadequate” in terms of fees and diversification opportunities in 62% of the cases. More generally, if interpersonal characteristics are not well diversified within a firm, or firm features play a big role in home bias (either through location effects, firm culture, industry, or plan features), the distribution of international allocations should remain relatively wide, compared to Fig. 1. Alternatively, if pure interpersonal characteristics are an important source of cross-individual variation in international allocations, aggregating over individuals in a firm is likely to eliminate much of the cross-sectional variation we 7

Also, Jonathan Reuter, in a discussion of our work, mentioned a median international allocation,

excluding target date fund allocations, of 21.2% using Brightscope data on 17,913 defined-contribution plans.



17

observe in Figure 1. Figure 3 reveals that 84.5% (69.90%) of average firm (relative) international allocations are in the 10%-25% (25%-50%) range, which is a much tighter distribution than in Figure 1. This suggests that personal characteristics may explain much of the observed interpersonal variation in international allocations. Insert Figure 4: Trends in International Diversification Finally, Figure 4 focuses on potential time effects in international diversification. Panel A shows that the quarterly average of idiv exhibit a marked upward trend (see also Coeurdaceur and Rey, 2013), roughly increasing from about 12% in 2005 to 22% in 2010 before dropping back to 18% in 2011, when European stock markets experienced a downturn following the flare-up of the sovereign debt crisis in August of that year. The figure also shows that the proportion of world markets accounted for by non-U.S. markets (dashed line) increased over time as well, moving from about 60% to 65% over the sample period. Thus, from the perspective of a simple World CAPM benchmark, international allocations should have increased over time. Alternatively, inertia coupled with different valuation changes for foreign versus domestic markets may also have caused individuals to become automatically more diversified over time.

Panel B

investigates this possibility by plotting the quarterly average of relidiv, which controls for the variation in the international equity market capitalization proportion, and shows a trend in international allocations over and above the trend in the underlying market capitalization benchmark. For this reason, we always include the benchmark foreign equity proportion as an independent variable in our regressions, and we will also verify whether relative returns in foreign versus domestic equity have a large effect on international allocations.

18

II. Personal Characteristics and International Diversification II.1 Trends, Age, and Cohorts Effects Trends The positive time trend we noted in Figure 4 can be due to a pure time effect, a positive cohort effect with older cohorts investing less in international stocks, a negative age effect coupled with a change in the age distribution, or some combination of the above. Ameriks and Zeldes (2004) show that these three effects, when modeled, as is usual, by dummy variables, are colinear and cannot be separately identified. Yet if the effects are persistent, identifying them will be important for predicting future trends in international diversification. In this section, we explore the time effects in international diversification. Table 2 reports some summary statistics on demographics and plan characteristics, while Table 3 reports the regression results. For each specification, we run three different panel OLS regressions, one with the listed independent variables, one controlling for firm fixed effects, and one controlling for ZIP code fixed effects (close to 30,000 different ZIP codes are represented in our sample). We cluster standard errors at the firm level to control for potential correlations among workers at the same firm due to, inter alia, changes in plan features, introduction of automatic enrollment (Madrian and Shea, 2001), or firm-level economic shocks. Firm-clustered standard errors are about 40 times larger than those in alternative specifications with no clustering and firm or ZIP code fixed effects.

We have verified that two important sources of this increase in the standard

errors are correlations between an individual’s allocation over time and the correlation



19

between individuals who join the firm at the same time, with such “tenure wave” clustering generating the largest increase in standard errors. Such workers may receive similar information regarding 401(k) plans, face similar investment options and return environments, and may even make personal contacts through investment information sessions that influence their investment decisions (Duflo et al., 2006). An important variable to control for in our regressions, and which may have bearing on the presence of trends, is the use of managed accounts and financial advice. First, when Financial Engines rolled out its managed accounts program, the pool of people in our sample may have changed toward more or less sophisticated people. The former is true if people consciously decide not to sign up for managed accounts when they are financially savvy enough to manage their 401(k) assets themselves. This may, in turn, account for an upward trend in international diversification.8 Alternatively, the people who sign up for managed accounts may be the ones who realize that financial advice should result in better diversified portfolios than what they could achieve without any help, leaving distracted and less sophisticated investors in the sample. While the percentage of individuals using managed accounts increased over time, we do not find the characteristics of the workers who use managed accounts—and which we drop from our sample—to be meaningfully different from those who do not. Panel C of Appendix Table 2 shows that the two groups are virtually identical in terms of observables such as age, salary, and account balance, although people using managed accounts have on average 4 fewer years of tenure than people who do not, since the introduction of such accounts is a recent phenomenon. Second, around 11% of the individuals in our sample 8



We thank Jonathan Reuter for pointing out this possibility.

20

sign up for financial advice, and approximately one third of those access it regularly. Because advice is an important determinant of international diversification, we dedicate a separate section to it (Section II.3). Third, because TDFs control the international asset allocation within their portfolios, we include a variable that represents the fraction of a person’s account balance that is invested in TDFs. The increasing popularity of TDFs may in fact have contributed to the increase in international allocations over time. As Table 2 shows, the average TDF allocation is 16.08%, with a number of plans not featuring TDFs at all, and some individuals investing their full balance in TDFs. Note that the major fund families (such as Vanguard and Fidelity) do not vary the international equity fraction over time or with age, so that we do not need to interact this variable with age. In addition, we control for the fraction of international assets relative to the world market capitalization—the idiv benchmark—which is one possible source of the trend in international diversification discussed in the previous section. We compute this fraction specifically for each person, based on the time at which the information on the allocations was drawn, and use it as an independent variable in all specifications. Our first regression in Table 3 simply adds a linear and quadratic trend to these two variables. All four independent variables have the expected coefficient. An additional 1% invested in TDFs is associated with a 0.08% higher international allocation. The effect is stable across specifications and highly statistically significant. Similarly, as the importance of the U.S. in the world markets decreases, we observe an increase in the average international equity allocation in our sample, with the effect varying between 0.19% and 0.21% per percent of international allocation. Trend coefficients, despite not



21

being statistically significant once we cluster standard errors, imply strong trends upward. The fitted temporal function generated by this trend specification is almost indistinguishable from the temporal function generated by an alternative specification with time dummies.9 We analyze a number of possible economic explanations for the increase in international allocations. First, we examine the role of cohort and/or age effects. Second, we examine the return experience effect described in Malmendier and Nagel (2011), return chasing, and simple valuation effects (foreign versus U.S. returns) coupled with inertia. Age and Cohort Effects Age and cohort results are reported in Columns (4) to (9) of Table 3. The cohort variable starts at 40 (for people born in 1940 or earlier) and ends at 90 (for people born in 1990 or later). Age is measured in years. Given that age and cohort variables are 99% negatively correlated, Table 3 reports regression results in which either the cohort or age variable is added to our baseline “trend” regression in linear form.10 While trend coefficients do not 9

While these time dummies are significant using OLS standard errors, many become insignificant when

clustered standard errors are used. We therefore use the economically equivalent quadratic trend specification. To further control for nonlinearities, we rerun the regressions in Table 3 and in the other Tables in the paper using a Tobit specification to account for the fact that 17% of the observations have zero international equity allocations. The results are robust, and a subset is reported in Appendix 5. 10

We also explored quadratic and more flexible dummy specifications for both cohort and age. We found

that the quadratic and dummy coefficients were mostly not statistically significant, and thus decided to proceed with a linear specification. We also ran a specification with firm-time fixed effects, in which the latter were either at the annual or quarterly level, and found that age and cohort effects were robust (available on request).



22

survive clustering of the standard errors, the age and cohort effects are always highly statistically significant. We find a positive cohort and a negative age effect. We postulate that the negative age effect is implausible on economic and statistical grounds. First, the age effect cannot really contribute to a general upward trend in international diversification unless the age distribution has shifted over time toward younger people. We examine the age distribution over time in our sample, and find it to be quite stable (results available on request). Second, the age effect implies that investors decrease their international allocations as they age, and that this decrease is counteracted by an overall trend toward more diversification. This interpretation of the results seems unlikely, although we cannot provide direct evidence that it is not occurring. Moreover, if the global trend does not persist, the graying of the population would imply that home bias, over the long run, would get worse in the aggregate. To test this directly but informally, we ran a regression of the change in idiv for each individual with multiple observations over the full sample onto a constant, the change in the benchmark idiv, and the change in the target date fund allocation. A negative age effect would tend to make the constant negative in such a regression. We obtain a highly significant positive constant. Of course, this may simply reflect the overall positive trend, but despite substantial crossheterogeneity in international diversification, only 26% of the population decreases its international diversification over time. A cohort effect is much more plausible, both economically and statistically. The simplest explanation for this is the ongoing globalization process that is familiarizing investors, particularly the younger generation, with global markets and global investments. If this is true, our results are potentially consistent with one of the most common findings in the



23

international literature regarding the effect of familiarity on home bias; we will return to this hypothesis when we investigate ZIP code effects. The potential long-run implications are important, as a sticky cohort effect would suggest that home bias will gradually go away. However, our cohort coefficient of 0.16-0.17 implies that an individual will increase her international allocation by about 1.6% over a decade, rendering the aggregate trend implications of the cohort effect rather modest. While the cohort variable explains about 10% of the total variation explained by all independent variables, the average cohort varies too little within our sample period to cause a marked increase in international allocations. The average cohort was (19)62 in 2006 and (19)65 in 2011, implying only a 0.5% aggregate increase in idiv over that period. Figure 5 shows international allocations by (coarse) cohorts, with people born before 1950, in the 1950s, 1960s, and 1970s, and after 1980. There is a monotonic relation from old (low idiv) to young (high idiv), but all cohorts also increase their international allocation over time. This result is very robust. We repeat the analysis by firm and make all the possible pairwise comparisons between the average idiv of a cohort and each of the older cohorts within each firm. There are 2,960 potential comparisons (10 comparisons per firm), but a few firms do not have observations for all the pairings. We find that the older cohort has higher international diversification in only 6.7% of the possible comparisons, and this figure drops to 4% if we consider only the cases in which the difference is significant at the 5% level. In contrast, in the 93.3% of the cases when the younger cohort is on average more diversified, the difference is statistically significant in 86% of the cases.11 What drives this overall diversification trend is unclear. 11 The youngest of the five cohorts is the most internationally diversified in 96% of the comparisons,

24

It may be due to the overall globalization phenomenon, which makes people more comfortable with international investing. Ongoing globalization may also affect international allocations by rendering the international investment opportunity set better over time, thereby enticing more international investment. Insert Figure 5 Cohorts and International Diversification Return-Sensitive Variables Another potential reason why cohorts matter is that investment behavior is affected by past return experiences. Malmendier and Nagel (2011) show that recent stock market experiences shape the risk taking and asset allocation of U.S. individuals. To examine this phenomenon, they create a weight function of past returns, depending on a parameter, λ, which can imply quite general weight patterns of past returns since birth. They find λ to be around 1.5, which means that recent returns are weighted more heavily than returns in the more distant past. Using SCF data and regressions that include, inter alia, age and time dummies, they show that this experience variable has a positive effect on stock market participation, risk tolerance, and the proportion of risky assets held by the individual. For our purposes, the relevant return is not the U.S. stock market return, but the difference between the foreign return, as measured as the MSCI international index whether or not we take the statistical significance of the difference into account. This corresponds to 273 of the 296 firms, as not all pairwise comparisons are available for all firms. The oldest cohort is the least diversified in 92% of the comparisons (266 firms; 96% if we only consider statistically significant differences), and in half of the cases in which the oldest cohort is not the least diversified, the second oldest is.



25

(excluding the U.S.), and the U.S. return. People who have firsthand experienced of poor international returns relative to the U.S. stock market (for example, during the Roaring ‘90s) may be more reluctant to invest abroad, and vice versa. The “MN (MalmendierNagel) experienced return” then becomes, in essence, a complex interaction of age, time effects, and past relative returns. We estimate λ together with the coefficient on the MN variable using nonlinear least squares. We run a number of preliminary regressions with fixed λ, using a relatively fine grid, and start the estimation at a λ that optimizes the R2 of the regression. We find the optimal λ to be 3.999 (see Table 4). This is substantially higher than the estimate in Malmendier and Nagel (2011) for the U.S. stock market, but still implies declining weights for relative returns. Because we only have international data from 1969 and there were virtually no international investments before 1980, a declining weight function seems the only plausible economic outcome. We find that the MN effect is statistically significant, but with a negative, rather than positive, coefficient, which is not consistent with the experience effect documented by Malmendier and Nagel.

To help interpret this finding, Figure 6 graphs the M-N

experienced return variable as a function of age for different points in time. Interestingly, the functions are mostly positive and decreasing with age; that is, younger people experienced more positive relative foreign returns, which may help explain the cohort effect we documented above. However, this effect is nonlinear and, depending on the year, from age 40 to 50 the effect becomes quite small (and even negative for 2005 and 2006, perhaps reflecting the experience of the ‘90s, when the U.S. stock market performed very well). For lower λ’s, we do find sometimes positive coefficients, but they are mostly not statistically significant. In contrast, the linear cohort effect remains

26

highly statistically significant and becomes larger. Since the MN and the cohort variable are 67% correlated, we also run a specification with the MN variable but excluding the cohort variable. In this specification, λ is estimated to be slightly above 1.0, but the MN variable still features a negative coefficient (results available on request). Regressions that replace cohort with age effects yield results similar to those reported in Table 4. Given these results, we conclude that the pure cohort variable is an easier to interpret and more robust determinant of variation in international allocations. We also examine an alternative specification of the MN variable by simply using the foreign return, rather than the foreign return minus the U.S. return. The idea of investors “chasing returns” in international markets is standard in the capital flow literature, going back to at least Bohn and Tesar (1996). When we run the nonlinear least squares model with this variant, we find that λ is equal to 1.00 and the MN variable has a positive effect on international allocations (see Table 4); that is, people who have experienced higher foreign returns allocate more internationally. However, the coefficient is no longer significant, while the cohort effect remains robust. An alternative explanation of the time variation in international diversification is inertia: Workers select an international allocation, perhaps when joining the firm, and never or rarely change it. If this is the case, the time variation in idiv should be partly explained by relative cumulative returns (foreign vs. U.S.) between the different records of account balances. We compute these individualized cumulative returns using daily MSCI returns. Column (3) of Table 4 shows that this variable has a negative sign and is not statistically significant. The introduction of firm or ZIP code fixed effects in Columns (4) and (5) does not change these conclusions. Note that the benchmark idiv variable, which remains

27

highly statistically significant, also partially reflects valuation changes. When we exclude this variable, the sign of the coefficient on the relative return duly becomes positive, but it is still not statistically significant. Finally, it is often suggested that in times of stock market crashes, investors become more risk averse and, at the same time, more home-biased. To test this conjecture, we rely on the indicator proposed by Baele et al. (2013), who use data on bond and stock returns to measure the occurrence of stress periods in which stock markets decline and liquid benchmark bonds increase in value. When we include in the regression the monthly incidence of these “Flights to Safety” days they identify for the U.S., we find that the coefficient on this variable is indeed negative in the simple OLS regression and the ZIP code fixed effect regression, but that it switches sign for the firm fixed effect regression and is not statistically significant. Given these nonrobust, hard to interpret, and/or insignificant results, we do not use any of the return-sensitive variables in the benchmark specification we take forward. II.2. The Effects of Income and Wealth We now examine whether income and wealth have an effect on international diversification. We have data on salary, account balances, and tenure at the firm. Since information on tenure is less complete, tenure is correlated with cohort, and account values may also largely reflect a combination of tenure and salary, we decided not to include tenure in the main specification. We also collected the median house value at the ZIP code level from Zillow, which is, for many households, perhaps the best indicator of overall wealth. We deflate all these variables using the 2005 CPI index.



28

Since the distributions of salary, account values, and house values are right-skewed, we use their natural logarithms in the regressions. We consider both linear and quadratic specifications. The quadratic term for house value is not statistically significant, but the quadratic terms for salary and account balances are; therefore, we include them in our final specification reported in Table 5.

We report again the usual three specifications,

but when ZIP code fixed effects are used, we must drop the house value as an independent variable because cross-sectional variation dominates time-series variation in house values in our sample. Note that most of our benchmark variables (% in TDF, international diversification benchmark, and cohort) maintain their sign and significance, with the coefficients on the TDF and cohort variables becoming slightly smaller, while the coefficient on the diversification benchmark increases substantially. Coefficients on salary, account balances, and house values are mostly statistically significant. The effect of house value on international diversification is positive. To get a sense of the economic magnitude, an increase in house value of $50,000 at the $200,000 average house value would generate roughly a 0.17% increase in idiv (the derivative with respect to house values for these magnitudes is the coefficient divided by 4). At the $58,000 average salary, an increase of $10,000 in salary would roughly generate a 0.69% increase in the international allocation coefficient (0.82% at the median salary of $47,625). For account balances, the negative quadratic effect makes international allocations a negative function of account balances. For the average account balance of $64,000, a $5,000 increase would generate only a 0.08% drop in international diversification (0.18% drop at the median account value of $23,434). Finally, note that



29

account balances and salary are positively correlated, so their joint effect may be somewhat smaller than the univariate effects. Because we lose many observations when we merge the data with the Zillow database, Columns (4) and (5) of Table 5 show the robustness of the results to an alternative data source for house values, namely, the Census Bureau’s American Community Survey. This survey provides the median house value per ZIP code over the 2008-2012 period. Hence, there is no panel available as with the Zillow database, but the coverage is larger. The table shows that the coefficients on account values and salaries are very close to those reported in the previous Columns, and the coefficient on house value now becomes somewhat higher, at 0.84 with firm fixed effects and 0.98 without, while retaining statistical significance. We conclude that differences in house values and account balances only generate economically small effects on international allocations, but we also detect more sizable salary effects. To check the robustness of this finding, we rerun the regressions in Table 5 for each firm separately. If we calculate the average house value for the workers in each firm and then estimate the effect on international diversification of increasing the average house value by $50,000, we find that the 25th to 75th percentile range of such effect varies between 0.07% and 0.41%, with a median of 0.20% and an average of 0.24%. If we calculate the average salary at each firm and increase it by $10,000, we find that the 25th to 75th percentile range of the effect of a higher salary varies between 0.12% and 1.69%, with a median of 0.68% and an average of 0.86%. Finally, a major potential source of heterogeneity in asset allocations is variation in risk aversion across individuals. There is, however, not an obvious link between risk aversion

30

and the optimal allocation to international assets in a portfolio. Under the CAPM benchmark, with a risk-free asset, optimality simply suggests holding the market portfolio, and our benchmark idiv is the optimal international equity portfolio. In a 401(k) context, in which shorting and leveraging are not possible, the optimal efficient risky frontier may have different international allocations for people with different risk tolerances. For example, high beta foreign investments (such as, currently, emerging markets) would be optimal for more risk-tolerant investors, and it is conceivable that higher income people are more risk tolerant. Therefore, individual- and ZIP code-level demographics might also capture variation in risk attitudes across individuals, although it is not clear which sign we should expect on many of the coefficients (see Harrison et al., 2004; Croson and Gneezy, 2009; and references therein). Finally, it is also possible that person-specific characteristics, experience, or behavioral biases account for differences in investment behavior (Cesarini et al., 2009; Campbell et al., 2014; Korniotis and Kumar, 2013). II.3 Access to Online Advice In this section we examine the effect on international allocations of signing up for financial advice. The proportion of individuals in our sample who signed an investor service agreement to receive online advice from Financial Engines is 11.27%. This proportion varies across firms, ranging from 0% to 55.1%, and also varies over time, ranging between 7% in 2005 and 10% in 2011.12 12

Over the sample period, on average, Financial Engines’ recommendations increased from roughly 20% to

40% in international equities, as a fraction of total equity exposure, although the actual percentages recommended vary substantially with age and risk profile.



31

Panel A of Table 6 compares the characteristics of employees who signed up for online advice with Financial Engines to those who did not. Online-advice employees are close to the same age—around 45—as those who do not sign up for advice and have similar house values, but have significantly higher salary, $73,095 vs. $56,420, and account values, $124,977 vs. $58,735. Interestingly, online-advice employees are also less likely to invest in Target Date Funds, having 7.74% of their portfolios in such funds vs. 16.83% for the non-advice employees. For those who signed up for financial advice, we also have information on the date of the last login on the online-advice website. Based on this, each time we observe an individual, we flag those whose last login is within a year as receiving non-stale advice. Among the individuals who signed up for advice, the advice is accessed within the past year on average 36.3% of the times. Summary statistics comparing those who have accessed the website within the past year and those who accessed it more than a year before the observation date indicate that the two groups are quite similar in terms of age, salary, account values, and reliance on TDFs (results available on request).13 Regressions in Panel B of Table 6 replicate Columns (2) and (3) of Table 5, controlling for whether the individual signed up for and has recently accessed online advice. Column (1) indicates that all else equal, individuals who signed up for advice have international 13

Notice that this is a very coarse measure of advice, as it flags as receiving advice those who have logged

onto the online advice website, without measuring whether they actually took advice, how long they stayed on the website, or which type of information they perused. Thus, the advice dummy coefficient measures the average correlation with international allocations across different intensities of advice. Comparison of the coefficients of stale vs. non-stale online advice suggests that a more precise measure of the degree of advice might generate a bigger effect.



32

allocations that are 4.27% higher than those who do not. This result is further confirmed in Column (2), in which we add a control for how recent the advice is: Someone who accessed the online advice website within the past year has, all else equal, international allocations that are 5.30% higher than those who did not sign up for advice and 2.25% higher than those who signed up but have not accessed the website recently. The regressions contain controls for firm fixed effects. Coefficients on the advice variables are similar in economic magnitude and significance when we instead control for ZIP code fixed effects (Columns (3) and (4)).14 Notice that we cannot tell from our data whether this different investing behavior is due to individuals who access online advice being different from otherwise similar people who do not, or to a direct effect of advice on international allocations. Finally, compared to Table 5, controlling for financial advice leaves the significance and economic magnitude of the coefficients on the other variables unchanged. Columns (5) and (6) examine the interactions between signing up for online advice and demographic variables. Coefficients indicate that signing up for advice is, all else equal, more strongly associated with higher international allocations for older cohorts—the cohorts that, based on the findings shown here and in Table 5, would otherwise invest less. This effect is statistically significant even after clustering standard errors at the firm level.

Similarly, among those who signed up for advice, all else equal, advice is

associated with increasing international allocations for individuals with lower account

14

Unreported results indicate that the effect of online advice and its recency are similar in regressions that

do not include firm or ZIP code fixed effects.



33

values. To the extent that older cohorts have higher account balances, these two effects may partially counteract each other. We also rerun the regressions in Tables 3, 4, and 5, dropping those individuals who signed up for advice at some point, and the results are robust (available on request). Given the importance of the advice variables and their interactions with demographics, we include them in the baseline specification that we use in the following sections. III. The Geography of Home Bias While personal characteristics and advice explain about 5.5% of the variation in international allocations (not reported), adding ZIP code fixed effects increases the adjusted R2 to significantly more than 9% (see Column (3) of Table 6, Panel B). To examine what these location effects reflect, we rerun our benchmark specification, including salary, account balance, advice variables, and their interactions, but excluding the house value variable, and extract the ZIP code fixed effects. We then run simple OLS regressions of these ZIP code effects onto a number of “locational” variables at either the ZIP code or state level. Our independent variables can be grouped into three broad themes: wealth, education, and familiarity/information. The first two are really personal characteristics that we can only measure at the ZIP code level. First, we include the ZIP code median house value in the regression. Because it substantially reduces our sample size (we only have house values for 12,446 ZIP codes), we typically run our specifications with and without this variable. Second, it is quite conceivable that education is correlated with financial savvy, and perhaps also helps alleviate any undue apprehension about foreign investments (Lusardi and Mitchell,



34

2007). The summary statistics in Table 7, Panel A, reveal that educational attainment displays substantial variation across ZIP codes. The 25-75th percentile range of the distribution is 56.9%-72.2% for the proportion of the population over 25 who has a high school diploma, 7.5%-18.6% for a bachelor’s degree, and 2.8%-9.8% for a master’s degree or higher. We also create a financial literacy variable, available at the state level, by computing average performance on the 5 financial knowledge questions in the National Financial Capability survey. Finally, most of our other variables can be related to the familiarity/information hypothesis. The first set concerns the percentage of the ZIP code population that is foreign born. If familiarity plays a large role in international allocations, it is conceivable that the presence of immigrants in a particular area directly or indirectly increases familiarity with foreign culture, products, and securities. Also, in the international literature, it is common to use distance from foreign markets as a control variable. Given the well-documented international foreign-investment biases toward nearby countries, we compute the distance to Toronto and to Mexico City and to London and Tokyo, the financial centers of the two largest investable equity markets outside the U.S. Finally, we measure whether the employee lives in a metropolitan area, large rural area, small rural area, or isolated area, using data from Rural Urban Commuting Area codes (RUCA). Familiarity relative to the foreign world can be also enhanced by working for a company that has a lot of business with foreign countries. We therefore include two measures of “trade openness”: the sum of imports and exports at the state level divided by state GDP and the level of exports divided by state GDP. Because the data on imports are less complete than those on exports, most of our analysis uses the export variable. Note that a



35

large literature in international finance, starting with Obstfeld and Rogoff (2000), links home bias in goods to home bias in assets through equilibrium models with transaction costs. However, Van Wincoop and Warnock (2010) show that such a link is empirically rather unlikely. Instead, our motivation for including these variables is based on a familiarity argument. Finally, a more direct measure of potential information flow would be the logarithm of the number of international phone minutes per year per state. Unfortunately, these data are not available, and therefore we use long-distance minutes as a proxy.15 To measure economic well-being, we include in the analysis GDP per capita and cumulative GDP growth over the five-year period preceding our sample and over the 2006-2011 period. These variables help mitigate concerns that any positive effect of the foreign-born population on international diversification is due to reverse causality: Richer areas or areas that are doing well economically are better diversified and, at the same time, attract more foreign immigrants. Before we consider the regression results, it is worth repeating that in our dataset location effects need not be highly correlated with firm effects. While it is true that many employees live close to where they work, our sample contains multiple firms with a multitude of branches that are quite spread out geographically. Table 7, Panel B, reports the regression coefficients for ZIP code effects extracted from a regression that includes our baseline specification (target date fund, benchmark idiv, trends, and cohort), plus the salary, account balance, advice variables, and their

15

Data are gathered from the FCC Statistical Trends in Telephony report; see Bekaert et al. (2014) for more

details. Data are spliced with data on interstate mobile phone minutes.



36

interactions.16 The table reports 6 different specifications, but three of those simply add the house value variable to an equivalent specification without the house value variable, which has many more observations. The first two specifications use coarse indicators of education (% bachelor’s degree or higher), immigration (total percentage born abroad), and distance (total distance). The third and fourth specifications are more granular with respect to education (high school, bachelor’s, or higher degree), origin of the foreignborn population, and the distance variable. In the fifth specification, we replace the Zillow house values with ones drawn from the census, which increases the number of observations considerably. Finally, in the last specification, we take the specification of Column (4) and replace the ratio of state exports to GDP with state openness. Key results can be easily summarized. First, we find a statistically significant effect of education and financial literacy levels on international allocations. A 1% increase in the percentage of people in the ZIP code who hold a bachelor’s degree or higher is associated with a 0.05% higher international diversification in that ZIP code. The effects are higher if we separately look at the people with high school, bachelor’s, and advanced degrees. To get a sense of the economic variation the coefficients imply, we evaluate the regression coefficients at the 25th-75th percentile range of the distribution reported in Panel A of Table 7. We use the coefficients from the base specification in Column (3) without the house values, but note that the coefficients are similar in the specifications with house values (Columns (2) and (4)-(6)). For a high school diploma, the international allocation is predicted to change over this range by about 0.48% (.0318*(72.2-56.9)); for 16

We verified that the results are robust to using ZIP code fixed effects derived from a regression with only

the baseline variables, which has slightly more observations.



37

a college degree by about 0.70% (.0634*11.1); and for a higher degree by about 1.06% (0.0633*16.8). Cumulatively, the effect is about 2.24%. We also examine financial literacy directly, and this variable has a coefficient that varies between 2.36 and 4.39, significant at the 1% level across all specifications. Going from poor to high financial literacy amounts to an increase in international diversification that varies between 0.47 and 0.88%, depending on the specification. We should also note that general education is already controlled for, so that improving financial literacy per se has the potential to increase international diversification outcomes. Second, we observe a “foreign-born” effect that is statistically significant at the 1% level in all specifications. Economically, the coefficient of 0.03 reflects a 0.20% increase in international diversification over the 0.4%-6.9% interquartile range for the foreign population variable. When we look at origin of immigration, we find that the variables are statistically significant for Latin American and European origin, but less consistently so for Asian origin. The strongest effect comes from the European origin, with a coefficient of 0.099. Third, overall distance has the expected negative effect on international diversification, but the effect is only statistically significant at the 10% level and loses significance once house values are included in the regression. When we split this variable into its components, the signs of the coefficients are not consistent across specifications. This is also true for the long-distance variable. Finally, we find a lower international diversification in both urban and larger rural areas (vs. isolated areas as the benchmark).17 17

Further analysis that includes only the rural-urban area dummies generates a positive and strongly

statistically significant coefficient for the urban dummy, suggesting that the reason for the difference is due



38

Fourth, GDP growth variables have mostly a significant positive effect on international diversification, and more so for lagged growth. In contrast, GDP per capita at the state level has a robust, statistically significant negative effect on international diversification. Fifth, trade openness generates strong and consistent positive effects on international allocations, both when measured using exports and when measured using both exports and imports for the shorter sample. The interquartile range for the exports variable, which is 2.9%, would induce an increase in international diversification that varies between 0.26% and 0.55% depending on the specification, while the interquartile range for state openness implies an increase in international diversification of about 0.53%. Finally, when other variables are controlled for, Zillow house values have a marginal or not statistically significant effect on ZIP code variation in international diversification. We conclude that there are relatively strong locational effects on international equity allocations that are related to education and financial literacy, immigration, GDP growth, and state trade openness. IV. Firm Characteristics and International Diversification Firm fixed effects substantially increase the adjusted R2 in the regressions we have run so far (see Panel B of Table 6). The reason may be twofold. First, the firm people work for affects their familiarity with and attitude toward foreign investments and goods, as well as the characteristics of their human capital. Second, the quality of the information provided to employees about available investment options and retirement investing might to other controls, such as foreign born, education, and financial literacy being strongly correlated with the urban dummy (results available on request).



39

vary across firms. In the worst-case scenario, a particular plan might not even offer an international mutual fund option. Alternatively, the options may be limited and/or have high fees, which would render international diversification less than optimal. We analyze the first reason in this section, and devote the next section to the effect of plan menus, and specifically the proportion of international funds offered by the company’s plans and their fees and quality. A firm’s culture or the firm’s activities may cause their employees to be more familiar and comfortable with investing internationally. Alternatively, working in a foreignowned firm or in an import/export-oriented industry may make employees less likely to invest internationally if they feel that, because of their jobs, their human capital is already largely linked to international markets. For this reason, we attempt to measure how “international” a firm is, either directly or indirectly.

We collect information from

CapitalIQ on the country of the ultimate parent of the company, for both private and public firms, and obtain information from Orbis on the fraction of foreign subsidiaries. Panel A of Table 8 shows that about 16% of the firms in our sample have a foreign parent, and 56% have at least one foreign subsidiary. Cross-firm variation in the fraction of foreign subsidiaries is vast, with an interquartile range from 0% to 69.1%. Additionally, we examine the openness, ((imports+exports)/output), of the industry the firm belongs to. We supplement this information with a number of variables that measure different firm characteristics, most of which do not have clear ex ante predictions regarding their effect on international diversification. First, we include a dummy variable to indicate whether the firm is private or publicly traded. Second, we include two measures of size: the



40

logarithm of the assets and the logarithm of the number of employees. We conjecture that employees at public and large firms may be more likely to be familiar with foreign investments, or they may have more elaborate and diverse 401(k) plans that offer more and better international options. We also use a leverage measure (debt/assets) and salesintensity measure (sales/assets). Third, we include measures of profitability (net income as a percent of assets) and investment intensity (capital expenditures as a percent of assets). Fourth, we include the logarithm of the age of the firm, where the logarithmic transformation is necessary, because some firms in our sample are very old. Table 8, Panel A, reports summary statistics on these firm characteristics. While we have panel data for some of these variables, most are very stable over time; therefore, we run our analysis on the firm fixed effects and simply average the independent variables. Again, we use the baseline specification, which includes target date fund allocation; idiv benchmark; trend variables: cohort variable; salary and wealth variables (account balances and house values); advice variables; and their interactions. We examine 6 different regression specifications. The first regression only includes controls for whether the company and its parent are public or private and the possible foreign location of its headquarters and subsidiaries. We find that individuals employed in private firms have significantly higher international equity allocations (Column (1)). Since 33% of the private firms in our sample have a public parent company—which is often foreign—we also control for whether the parent is public or private and for the presence of foreign headquarters. The table shows that while working for a private firm is associated with 3.92% higher international allocations, the effect decreases to only 1.68% when there is a public parent company. The coefficient on the Foreign Headquarters



41

dummy is not significantly different from zero, possibly due to the opposing effects of familiarity and human capital considerations described at the beginning of this section. In the remaining regressions, we add firm characteristics such as size, age, profitability, etc., as well as a dummy for whether the firm has foreign subsidiaries. Notice that our sample now loses about 170 firms, for which not all the data are available; most of these firms are private. The regression in Column (2) indicates that having foreign subsidiaries is associated with a 3.59% higher international allocation, and that the effect is statistically significant. 18 Among the other firm characteristics, only profitability is significantly correlated with international allocations, with a negative effect amounting to 0.17% per percent of profitability. One possibility is that workers in profitable firms invest disproportionally in company stock, crowding out international investments. To examine this substitution effect further, we calculate the aggregate allocation to company stock at the firm level. That is, we take the last observation on company allocations per individual in each year and multiply this allocation by total account value to obtain a dollar allocation and aggregate this over each firm-year. We then match firm-year aggregate company stock allocations to firm-year profitability, leading to 513 observations for profitability and company stock allocations. We find a positive but small correlation between the two variables, at 10.9%, which is significant at the 1% level. In Column (3) we add industry effects and in (4) industry openness, to control for the international nature of the firm the individual works for. Private and foreign subsidiaries dummies and the profitability measure remain significant, both statistically and economically, while the new variables are not statistically significant; however, the 18



Controlling for the fraction of foreign subsidiaries instead of the dummy yields similar results.

42

industry dummies increase the R2 substantially. Finally, controlling only for firm characteristics and not for ownership or foreign subsidiaries (Column (5)) does not change the results. IV. Plan Quality and Menu Design One important determinant of the degree of international diversification in portfolios, and more generally of the allocation to different asset classes, is the availability of international investment options and their quality and fees. To examine this issue, we take two different approaches. First, we replicate the results in Tables 3, 5, and 6, controlling for fixed effects based on the quarter the individual joined the firm interacted with the firm’s identity and further interacted with the quarter-year of observation. The rationale behind this approach is to control for the host of conditions that a group of workers who join a given firm in the same quarter faces at the time they join and over time (for example, automatic enrollment; training sessions; social interaction; investment options available, their quality and fees, and their evolution over time). Panel A of Table 9 reports the results and shows that the coefficients are robust across specifications and, in most cases, only slightly smaller than those obtained previously. The only exception is the cohort effect, which is now one third of its former size, as the firm/date/quarter the worker joined fixed effects are correlated with the cohort effect in the data. In addition, the R2 is now more than 13% across specifications. Second, we replicate the main results in Tables 3, 5, and 6 by controlling directly for the fraction and quality of the international funds offered by the plans, based on information obtained from a snapshot of the plans offered by the firms in our sample in 2012. As



43

such, we capture the most recent features of the plans and, to the extent that there is correlation over time, historical features as well. For each plan, we calculate the ratio of the number of international over domestic equity funds, the ratios of median international over domestic turnover and fund age, a dummy for plans with high fees compared to similar plans, the difference between the median historical alphas of international and domestic funds, the difference between the median expense ratio of international equity funds and domestic ones, and total plan size.19 The last rows of Panel A, Table 8 report the summary statistics for these plan characteristics and show substantial variation across plans. On average, 20% of the equity funds in the plans in our sample are international; their turnover and age are similar to, albeit slightly lower than, the domestic funds in the same plan; and they have slightly lower historical alphas, both in terms of means and medians.20 As expected, the international funds tend to have higher expense ratios than the domestic funds, but the differences are small, with an average difference of 0.05% and a median difference of 0.03%21.

19

Note that one quality dimension we do not investigate is potential heterogeneity in the target date funds

offered by the plans. Balduzzi and Reuter (2015) actually find large heterogeneity in performance, asset allocation, market timing, and security selection among target date funds with the same target date across fund families. 20

Alphas are calculated relative to the style return produced by Financial Engines’ style analysis with 15

asset classes. 21

According to the Morningstar Fee Trend Report the average difference is 0.22% in 2012 (0.20% using

asset-weighted averages). This is much higher than in our 401(k) data, perhaps indicating preferable share classes as a result of sponsor negotiating power.



44

Panel B of Table 9 reports the coefficients on the plan features. Column (1) replicates Column (4) of Table 3, while the following four Columns replicate Columns (2) and (5) of Tables 5 and 6, respectively. The full version of the table is available in the Appendix. Shifting the fraction of international funds over the interquartile range, from 16.7% to 25.6%, is enough to increase international allocations by between 1.57% and 1.81%, depending on the specification. When we instead consider the 90% range, the effect varies between 4.03% and 4.65%, which is similar in magnitude to that of having an individual who goes from no online advice to non-stale online advice. The coefficient is statistically significant across specifications. The effect of fees is also highly statistically significant. Lowering expense ratios from the 75th to the 25th percentile of the distribution is associated with international allocations that are approximately 0.75% higher. Given the dispersion in fees, the effect jumps to 3.12% if we instead consider the 95th-5th range. Our results confirm country-level findings on transaction costs and international diversification (French and Poterba, 1991; Cooper and Kaplanis, 1994; Jeske, 2001). Surprisingly, poor plan quality—as reflected in the relatively high expense ratios of funds in the plan compared to the universe of similar funds—has a positive and significant effect.22 Finally, adding such plan features to the regressions does not change the overall magnitude or statistical significance of the other coefficients when compared to Tables 3, 5, and 6.

22

Further analysis suggests that this could be due to companies with foreign headquarters having higher fee

plans and, at the same time, employees investing more internationally, since we find a positive correlation between the high fee dummy and the foreign headquarter dummy.



45

To the extent that plan features are determined by employers and are not exclusively the result of employees’ demands, the findings above indicate that improving the quality of international investment options offered by 401(k) plans, in terms of the number of funds available and their fees, should generate more investment in international equity. Lastly, we add these plan characteristics to the analysis of the firm fixed effects presented in Panel B of Table 8. Column (6) shows that these variables contribute substantially to explain the fixed effects, as the R2 increases from 16.5% to 29%. Among the plan features, expense ratios and plan quality, in terms of fees relative to the universe of similar funds, are the only ones that are statistically significant. VI. Robustness Checks While we have already reported on a number of robustness checks along the way, here we specifically focus on the problem posed by the fact that our data represent one 401(k) account per person, which may not be representative of the individual’s full portfolio. To investigate this issue, we focus on subsamples of individuals for whom there is a high chance that their wealth is dominated by their 401(k) account, and that this 401(k) account is their only account. Of course, our selection criteria will use variables that are themselves correlated with international diversification. While this is not desirable, it would render robust results all the more surprising. Our first criterion simply uses tenure and age, and is based on the fact that relatively old workers with relatively short tenure at the firm are more likely to already have a 401(k) account from a previous employer or to have an IRA account (see Table 10 and the Appendix for details). We also create a subsample based on salary and account value,



46

excluding individuals with a salary of more than $100,000 or an account balance of more than $200,000. Such individuals are likely to have substantial taxable assets, which renders their 401(k) account less representative of their overall allocation. Finally, we create a subsample that combines both criteria. In Table 10, we show these results in Columns (2) through (4), focusing on the benchmark specification with only the target date fund variable, idiv benchmark, trends, and cohort. In Panel A of Appendix Table 4, we add salary, account balances, and house values, while in Panel B of Appendix Table 4 we add the advice variables and their interactions with cohort, salary, and account values. All regressions include firm fixed effects, although the results below are robust to other specifications. The target date fund variable, international diversification benchmark, and cohort effect remain statistically significant and of similar magnitude in all the subsample specifications, while trend coefficients are less robust. Similarly, salary and house value effects are robust, and the advice variable and its interactions with cohort and account values maintain their economic magnitude and statistical significance, although the interaction with salary is less robust. We also investigate our accounts’ bond allocations. A high allocation to bonds within a 401(k) account (which is tax-advantaged) may indicate an asset location strategy and suggest a sizable taxable portfolio. The mean allocation to bonds (conditional on equity market participation) is 18.64%, with the 75% range going from 4% to 29%. As we explained before, our focus on idiv (foreign equity over total equity) implies that high bond allocations may not necessarily be a problem. However, to increase the representativeness of the sample, we also investigate a sample that excludes accounts



47

with bond allocations of more than 50%. Again, Column (5) in Table 10 (and Panels A and B of Appendix Table 4) shows that the results are quite robust. The last Column of Table 10 (specification (6)) reports results in which we change the left-hand-side variable to the proportion of overall assets invested in international equities, which would be the natural outcome variable in an allocation model. Recall that we do not observe the allocation to international bonds, although we surmise that it is relatively small. The main results remain largely intact. Finally, since we do not observe our investors’ actual holdings, it is possible that some of them may invest internationally by holding U.S. portfolios (stocks) that have more exposure to international factors, e.g., multinationals (Cai and Warnock, 2012). Both old research by Jaquillat and Solnik (1978) and newer results by Rowland and Tesar (2004) suggest that multinationals alone are not sufficient to span the international diversification benefits from investing in local foreign companies. Since we do not have information on multinational investments—and thus cannot see directly whether investors use multinational companies as a substitute for international investments—we exploit the fact that our dataset does split the U.S. equity portfolio between small and large companies. Given that multinational companies tend to be large, we calculate the correlation between the international equity allocation and the allocation to large U.S. equities in the portfolios of employees in our sample, and we actually find it to be positive at 12.6% and increasing over time. It is therefore unlikely that investors use large U.S. companies as substitutes for international diversification. VII. Conclusions



48

We have examined the international equity allocations of 3.8 million individuals in 296 401(k) plans over the 2005-2011 period. A striking feature of the data is the very large cross-individual variation in these allocations, with nonnegligible fractions of individuals allocating 0% to international equities and, simultaneously, a minority allocating more than 75%. We examine a number of sources of variation in these allocations: pure temporal trends, personal characteristics such as salary and wealth, access to financial advice, ZIP code effects, and firm effects, including the quality and number of international funds offered by the plans. We find a strong cohort effect, with younger cohorts investing more internationally, but each cohort also investing more internationally over time. Access to online advice and its recency are, all else equal, associated with sizably higher international allocations. The fraction of international funds offered by the plan and their relative fees compared to the domestic funds in the plan have effects of similar size. In addition, we find a positive salary and a negative account balance coefficient, but only the salary effect is economically meaningful. Level of education measured at the ZIP code level has a positive effect on international diversification, as does financial literacy. The fraction of foreign-born population at the ZIP code level is also associated with higher international allocations. In addition, individuals who live in states with more exports, or work for companies that are private, have foreign subsidiaries, or are less profitable, have higher international allocations. The cohort effect, coupled with more access to financial advice, education, and better international fund choices, might cause the home-bias phenomenon to slowly disappear over time. These results point to a potentially big role for public policy in correcting individual investment mistakes and improving retirement outcomes.



49

A number of our results are also consistent with the familiarity hypothesis stressed in international finance literature; this includes the cohort effect, which may stem from globalization rendering younger people more comfortable with international investing. However, there are clearly other forces at work as well, and we only explain a small part of the total cross-individual variation. Because we only have data on 401(k) allocations, which for many individuals may not represent their full investment portfolio, it is conceivable that some people who underinvest in international equity in their 401(k) plans have international allocations elsewhere. Taking taxes into account, asset location optimization would suggest skewing the 401(k) portfolio towards bonds. We partially accommodate this critique by focusing on the relative equity allocation. In addition, we have examined various subsamples that minimize the incomplete portfolio problem, excluding people who have very short tenure but are older, and/or low account balances and/or salary, in order to focus on people for whom the 401(k) account we observe is likely the biggest—if not the only—part of their financial portfolio. We also investigate a sample that excludes accounts with excessive bond allocations, which may also suggest an asset location strategy. Our results remain robust in all of these subsamples. At this point, we have studied international equity allocation conditional on equity market participation. It might also be interesting to study the decision to participate in the international equity market by itself, as international allocations are 0 in 17% of the observations. This behavior is only partly correlated with general stock market nonparticipation, and might be heavily correlated with other behavioral investment



50

biases/mistakes, such as excessive allocations to money market instruments and/or company stock. We defer analyzing this to future work. Quantifying whether the effects we document are economically important requires an asset allocation model and many auxiliary assumptions, which is beyond the scope of this paper. However, our evidence on the dispersion of international equity allocations across U.S. individuals, and on the variables that are correlated to it, could serve as a starting point for future theoretical work. Our findings also have important implications for the international finance literature on home bias. First, many of our results confirm the importance of familiarity and information-flow stories (Andrade and Chhaochharia, 2010; Van Nieuwerburgh and Veldkamp, 2009), which should be researched in more detail. Second, the large cross-individual variation linked to cohorts, education, financial literacy, and access to financial advice should lead to additional analysis of cross-country home bias that focuses on heterogeneity in investor population, which hitherto has not been fully examined.



51

References Ahearne, A. G., Griever, W. L., Warnock, F. E., 2004. Information costs and home bias: an analysis of U.S. holdings of foreign equities. Journal of International Economics 62, 313-336. Ameriks, J., Zeldes, S., 2004. How Do Household Portfolio Shares Vary with Age? Unpublished working paper, Columbia University. Andrade, S., Chhaochharia, V., 2010. Information immobility and foreign portfolio investment. Review of Financial Studies 23, 2429-2463. Aviat, A., Coeurdacier, N,. 2007. The geography of trade in goods and asset holdings. Journal of International Economics 71, 22–51. Baele, L., Bekaert, G., Inghelbrecht, K., Wei, M., 2013. Flights to Safety. Unpublished working paper, Columbia University. Balduzzi, P., Reuter, J., 2015. Heterogeneity in Target-Date-Funds: Optimal Risk Taking or Risk Matching? Unpublished working paper, Boston College. Becker, S.O., Hoffmann, M., 2010. Equity fund ownership and the cross-regional diversification of household risk. Journal of Banking & Finance 34 (1), 90-102. Bekaert, G., Siegel, S., Wang, X., 2013. Home Bias Revisited. Unpublished working paper, Columbia University. Bekaert, G., Harvey, C. R., Lundblad, C., Siegel, S., 2011. What Segments Equity Markets? Review of Financial Studies 24 (12), 3847-3890. Bekaert, G., Harvey, C. R., Lundblad, C., Siegel, S., 2014. Stock Market Valuation across US States. Unpublished working paper, Columbia University.

52

Bohn, H., Tesar, L., 1996. U.S. Equity investment in foreign markets: portfolio rebalancing or return chasing? American Economic Review: Papers and Proceedings 86, 77-81. Brennan, M. J., Cao, H. H., 1997. International portfolio investment flows. Journal of Finance 52, 1851-1880. Brown, J. R., Pollet, J. M., Weisbenner, S.J., 2015. The In-State Equity Bias of State Pension Plans. Unpublished working paper, UIUC College of Business. Cai, F., Warnock, F.E., 2012. Foreign exposure through domestic equities. Finance Research Letters, 8–20. Calvet, L., Campbell, J., Sodini, P., 2007. Down or Out: Assessing the Welfare Cost of Household Investment Mistakes. Journal of Political Economy 115(5), 707-747. Calvet, L., Campbell, J., Sodini, P., 2009. Fight or Flight? Portfolio Rebalancing by Individual Investors. Quarterly Journal of Economics 124, 301-348. Calvet, L., Campbell, J., Sodini, P., 2009. Measuring the Financial Sophistication of Households. American Economic Review, Papers & Proceedings 99(2), 393-398. Campbell, J., Ramadorai, T., Ranish, B., 2014. Getting Better: Learning to Invest in an Emerging Stock Market. Unpublished working paper, Harvard University. Cesarini, D., Dawes, C. T., Johannesson, M., Lichtenstein, P., Wallace, B., 2009. Genetic variation in preferences for giving and risk taking. Quarterly Journal of Economics 124, 809-842.



53

Chan, K., Covrig, V., Ng, L., 2005. What determines the domestic bias and foreign bias? Evidence from mutual fund equity allocations worldwide. The Journal of Finance 60, 1495-1534. Coeurdacier, N., Rey, H., 2012. Home Bias in Open Economy Financial Macroeconomics. Journal of Economic Literature 51, 1, 63-115. Cooper, J., Kaplanis, E., 1994. Home bias in Equity portfolios, inflation hedging and international capital market equilibrium. Review of Financial Studies 7 (1), 45-60. Coval, J., Moskowitz, T., 1999. Home Bias at Home: Local Equity Preferences in Domestic Portfolios. The Journal of Finance 54, 2045- 2073. Coval, J., Moskowitz, T., 2001. The Geography of Investment: Informed Trading and Asset Prices. Journal of Political Economy 109, 811-841. Croson, R., Gneezy, U., 2009. Gender Differences in Preferences. Journal of Economic Literature 47, 2, 448-474. Curcuru, S., Heaton, J., Lucas, D., Moore, D., 2004. Heterogeneity and Portfolio Choice: Theory and Evidence. In: Ait0Sahalia, Y., Hansen, L.P., (Eds.), Handbook of Financial Econometrics, Vol. 1, North-Holland, Amsterdam, pp. 337-382. Dahlquist, M., Pinkowitz, L., Stulz, R. M., Williamson, R., 2003. Corporate governance and home bias. Journal of Financial and Quantitative Analysis 38, 87-110. Duflo, E., Gale, W., Liebman, J., Orszag, P., Saez, E., 2006. Saving Incentives for Lowand Middle-Income Families: Evidence from a Field Experiment with H&R Block. Quarterly Journal of Economics 121(4), 1311-1346.



54

Elton, E., Gruber, M., Blake, C., 2006. The Adequacy of Investment Choices Offered By 401(k) Plans. Journal of Public Economics 90 (6-7), 1299-1314. Fidora, M., Fratzscher, M., Thimann, C., 2007. Home bias in global bond and equity markets: The role of real exchange rate volatility. Journal of International Money and Finance 26, 631-655. French, K., Poterba, J., 1991. International diversification and international equity markets. American Economic Review Papers & Proceedings 81(2), 222-226. Glassman, D., Riddick, L., 2001. What causes home asset bias and how should it be measured? Journal of Empirical Finance 8, 35-54. Graham, J., Harvey, C., Huang, H., 2009. Investor Competence, Trading Frequency and Home Bias. Management Science 55(7), 1094-1106. Grinblatt, M., Keloharju, M., 2001. How Distance, Language and Culture Influence Stock Holdings and Trades. The Journal of Finance 56 (3), 1053-1073. Grinblatt, M., Keloharju, M., 2001. What Makes Investors Trade? The Journal of Finance 56 (2), 589-616. Harrison, G., Lau, M., Rutstrom, E., 2004. Estimating Risk Attitudes in Denmark: A Field Experiment. Scandinavian Journal of Economics 109(2), 341-368. Huang, J., 2008. Taxable and Tax Deferred Investing: An Arbitrage Approach. Review of Financial Studies 21, 2173-2207. Huberman, G., 2001. Familiarity breeds investment. Review of Financial Studies 14, 659-680.



55

Jacquillat, B., Solnik, B., 1978. Multinationals are poor tools for diversification. Journal of Portfolio Management 4, 8-12. Jeske K., 2001. Equity Home Bias: Can Information Cost Explain the Puzzle? Federal Reserve Bank of Atlanta Economic Review 86, 3, 31-42. Kang, JK, Stulz, R., 1997. Why is there a home bias? An analysis of foreign portfolio equity ownership in Japan. Journal of Financial Economics 46(1), 3-28. Karlsson, A., Norden, L., 2007. Home sweet home: Home bias and international diversification among individual investors. Journal of Banking and Finance 31, 2, 317333. Kho, B., Stulz, R., Warnock, F.E., 2009. Financial Globalization, Governance and the Evolution of the Home Bias. Journal of Accounting Research 47(2), 597-635. Korniotis , G., Kumar, A., 2013. Do Portfolio Distortions Reflect Superior Information or Psychological Biases? Journal of Financial and Quantitative Analysis 48(1), 1-45. Lewis, K, 1999. Trying to Explain Home Bias in Equities and Consumption. Journal of Economic Literature 37(2), 571-608. Lusardi, A., Mitchell, O., 2007. Baby Boomer retirement security: The roles of planning, financial literacy, and housing wealth. Journal of Monetary Economics 54(1), 205-224. Madrian, B., Shea, D. 2001. The Power of Suggestion: Inertia in 401(k) Participation and Savings Behavior. Quarterly Journal of Economics 116(4), 1149-1187. Malmendier, U., Nagel, S., 2011. Depression Babies: Do Macroeconomics Experiences Affect Risk-Taking? Quarterly Journal of Economics 126(1), 373-416.



56

Norden, L., 2010. Individual Home Bias, Portfolio Churning and Performance. European Journal of Finance 16(4), p. 329-351. Obstfeld, M., Rogoff, K., 2000. The Six Major Puzzles in International Macroeconomics: Is There a Common Cause? In: Bernanke, B. S., Rogoff, K., (Eds.), NBER Macroeconomics Annual. Portes, R., Rey, H., 2005. The determinant of cross-border equity flows. Journal of International Economics 65, 269-296. Rowland, P., Tesar, L., 2004. Multinationals and the Gains from International Diversification. Review of Economic Dynamics 7, 789-826. Sercu, P., Vanpee, R., 2012. The home bias puzzle in equity portfolios. International Finance: A Survey, 310-331. Sharpe, W.F., 1992. Asset allocation: Management style and performance measurement. Journal of Portfolio Management, 7-19. Van Nieuwerburgh, S., Veldkamp, L., 2009. Information Immobility and the Home Bias Puzzle. Journal of Finance 64(3), 1187‐1215. Van Wincoop, E., Warnock, F.E., 2010. Is Home Bias in Assets Related to Home Bias in Goods? Journal of International Money and Finance 29, 1108-1123.



57

Figure 1 International Diversification across Individuals Panel A shows the distribution of international equity allocations as a percent of total equity allocations across individuals’ 401(k) portfolios. Panel B shows the distribution of this ratio relative to an international diversification benchmark calculated as the ratio of the MSCI Market Cap All Countries ex-US Index to MSCI Market Cap All Countries Index. The sample in both figures is restricted to stock market participants (individuals with positive total equity allocations). All variables are defined in the Appendix.

Panel A – International Diversification 40 35 % of popula)on

30 25 20 15 10 5 0 0

(0,10]

(10,25]

(25,50]

(50,75]

>75

Intena)onal Equity/Total Equity

Panel B – Over and Under International Diversification 45 40 % of popula)on

35 30 25 20 15 10 5 0 0

(0,10]

(10,25] (25,50] (50,90] (90,110] (110,125] (125,150] >150 Interna)onal diversifica)on rela)ve to benchmark





58

Figure 2 International Diversification across States Figure 2 shows maps with the distribution of international equity allocations as a percent of total equity allocations across U.S. states at different points in time. The sample is restricted to stock market participants (individuals with positive total equity allocations).

Panel A - International Diversification across States in 2007

Panel B - International Diversification across States in 2010



59

Figure 3 International Diversification across Firms Panel A shows the distribution of the average across firms of idivi,t, the international equity allocation as a percent of total equity allocation. Panel B shows the distribution of the average across firms of relidivi,t, calculated by dividing idivi,t by an international diversification benchmark equal to the ratio of the MSCI Market Cap All Countries ex-US Index to MSCI Market Cap All Countries Index. The sample is restricted to stock market participants (individuals with positive total equity allocations). All variables are defined in the Appendix.

Panel A – International Diversification 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 0

(0,10]

(10,25]

(25,50]

(50,75]

Average Idiv by Firm

Panel B – International Diversification Relative to the MSCI Benchmark 80% 70% 60% 50% 40% 30% 20% 10% 0% 0

(0,10] (10,25] (25,50] (50,90] (90,110] (110,125] (125,150] >150 Average relidiv by Firm





60

Figure 4 International Diversification over Time Panel A plots quarterly averages of idivi,t (solid line), the international equity allocation as a percent of the total equity allocation, against the international diversification benchmark (dashed line), defined as the ratio of the MSCI Market Cap All Countries ex-US Index to MSCI Market Cap All Countries Index. Panel B plots quarterly averages of relative international diversification (the ratio of idivi,t to the benchmark). The sample is restricted to stock market participants (individuals with positive total equity allocations). All variables are defined in the Appendix.

Panel A: Trends in International Diversification and the MSCI Benchmark 67

24

66

22

65

20

64

18

63

16

62

14

61

Benchmark

2011q4

2011q3

2011q2

2011q1

2010q4

2010q3

2010q2

2010q1

2009q4

2009q3

2009q2

2009q1

2008q4

2008q3

2008q2

2008q1

2007q4

2007q3

2007q2

2007q1

10

2006q4

59

2006q3

12 2006q2

60

InternaFonal DiversificaFon (RHS)

Panel B: Trends in Relative International Diversification 34 32 30 28 26 24 22

2011Q4

2011Q3

2011Q2

2011Q1

2010Q4

2010Q3

2010Q2

2010Q1

2009Q4

2009Q3

2009Q2

2009Q1

2008Q4

2008Q3

2008Q2

2008Q1

2007Q4

2007Q3

2007Q2

2007Q1

2006Q4

2006Q3

18

2006Q2

20

RelaFve InternaFonal DiversificaFon



61

Figure 5 Cohorts and International Diversification

30 25 20 15

1980

Figure 6 Malmendier and Nagel Experienced Returns

Experienced Returns with λ = 3.999

Following Malmendier and Nagel (2011), the experienced returns variable is the weighted average of past returns with weights that depend on an individual's age at time t, how many years ago the return was realized and a parameter λ that controls for the shape of the weighting function. This paper defines past returns using international stock returns in excess of U.S. stock returns. The figure below shows experienced returns for λ = 3.999. 8%

6% 4% 2% 0% -2% 16

21

26

2005



31

36 2006

41

46

51

2007

56

61 66 Age 2008

71

76 2009

81

86

91

2010

96 101 106 2011



62

Table 1 International Under-Diversification in the US This table reports statistics for the degree of international diversification, i.e. the international equity allocations as a percent of total stock allocations in individuals’ 401(k) portfolios. Firms (states) are ranked according to the average international diversification: diversified firms (state) represent the 5 firms (1 state) with the highest average diversification and under-diversified firms represent the bottom 5 firms (1 state). This subsample is then split into older people (born in 1960 or earlier) and younger people (born 1980 or later). Finally, within each cohort, individuals are split in terciles (low salary, intermediate salary and high salary). The reported numbers are the average international diversification for each subset.

Diversified Firms Cohort 1960 Low salary Intermediate salary High salary Cohort 1980 Low salary Intermediate salary High salary

Under-diversified Firms

33.1

3.70

30.7

4.86

33.7

6.76

39.0

10.2

36.3

11.8

37.4

13.7

Diversified State

Underdiversified State

22.2

13.3

19.7

11.2

19.1

13.6

31.2

21.0

27.7

19.1

25.8

19.3

Cohort 1960 Low salary Intermediate salary High salary Cohort 1980 Low salary Intermediate salary High salary

Table 2 Summary Statistics for Stock Market Participants This table reports the mean, median, std dev, 25th and 75th percentiles and number of observations for the individual level data. The sample includes individuals with a positive equity allocation in their 401(k) portfolio. All variables are defined in the Appendix. The sample period is 2005 to 2011.



Variable

Mean

Median

Std Dev

p25

p75

# Obs

International Div

17.736

17.073

15.959

3.846

26.263

17,082,302

% in Target Date Fund

16.077

0.000

33.774

0.000

0.000

17,082,302

International Div Benchmk

64.368

64.856

1.462

63.304

65.428

17,082,302

Cohort

(19)63

(19)63

11.683

55.000

72.000

17,082,302

Age

45.480

46.000

11.667

37.000

54.000

17,082,302

Annual Salary

58,017

47,625

47,772

20.136

65.463

13,105,091

Total Account Value

63,972

23,434

113,906

5.714

75.250

17,054,517

House Value

244,143

188,133

188,858

126,300

299,555

13,984,030

MN Experienced Returns

0.96

0.819

0.837

0.396

1.287

17,082,302

MN Return Chasing

8.692

8.748

1.267

7.986

9.954

17,082,302

Relative Returns

0.001

0.000

0.061

-0.039

0.031

17,082,302

Flight to Safety Dummy

0.062

0.000

0.119

0.000

0.091

17,082,302

Tenure

12.82

10.21

10.35

4.37

19.08

9,170,900

Advice Dummy

0.088

0

0.284

0

0

17,082,302

Not Stale Advice Dummy

0.032

0

0.176

0

0

17,082,302

63

Table 3 On Time, Cohorts and Age This table reports the results for individual level regressions of international diversification on a quadratic time trend, birth year cohort and age, all controlling for the percent invested in a target date fund and the international diversification benchmark. Columns (2), (5) and (8) control for firm fixed effects, while Columns (3), (6) and (9) control for Zip code fixed effects. All variables are defined in the Appendix. T-statistics are in brackets. The superscript *** denotes significance at the 1% level, ** at the 5% level and * at the 10% level. Standard errors are clustered at the firm level. The sample period is 2005 to 2011.

Variables % in Target Date Fund Intl Div. Benchmark Trend Trend2

(1) Idiv 0.0796*** [9.061] 0.196 [1.315] 0.0254 [0.154] 0.00768 [1.374]

(2) Idiv 0.0697*** [13.59] 0.214*** [2.719] 0.0276 [0.235] 0.00435 [0.961]

(3) Idiv 0.0783*** [10.83] 0.193 [1.559] -0.0442 [-0.305] 0.00943* [1.833]

Cohort

(4) Idiv 0.0684*** [7.754] 0.210 [1.428] 0.0573 [0.358] 0.00557 [1.054] 0.167*** [11.19]

(5) Idiv 0.0590*** [11.79] 0.214*** [2.759] 0.0395 [0.348] 0.00331 [0.753] 0.158*** [15.41]

(6) Idiv 0.0682*** [9.376] 0.205* [1.677] -0.0158 [-0.111] 0.00751 [1.535] 0.161*** [12.88]

Age Constant

Observations Adjusted R-squared Firm Fixed Effects Zip Code Fixed Effects



(7) Idiv 0.0682*** [7.743] 0.208 [1.417] 0.102 [0.636] 0.00552 [1.045]

(8) Idiv 0.0588*** [11.80] 0.211*** [2.680] 0.0808 [0.708] 0.00329 [0.746]

(9) Idiv 0.0680*** [9.369] 0.203* [1.659] 0.0268 [0.189] 0.00746 [1.525]

-0.159*** [-15.25] 8.402 [1.624]

-0.163*** [-13.13] 8.693 [1.199]

1.532 [0.178]

1.302 [0.258]

2.368 [0.328]

-9.752 [-1.152]

-8.450* [-1.736]

-8.427 [-1.190]

-0.170*** [-11.45] 8.040 [0.943]

17,082,302 0.039 N

17,082,302 0.120 Y

17,063,721 0.073 N

17,082,302 0.053 N

17,082,302 0.131 Y

17,063,721 0.086 N

17,082,302 0.054 N

17,082,302 0.132 Y

17,063,721 0.086 N

N

N

Y

N

N

Y

N

N

Y

64

Table 4 The Effect of Return Sensitive Variables This table reports the regressions in Columns (4)-(6) of Table 3 Panel A adding return sensitive variables (experienced returns, intnl stock returns relative to U.S. returns, and a flight to safety dummy). The specifications in Columns (1) and (2) were run using non-linear least squares, where lambda measuring how the effect of past returns decay with time. All variables are defined in the Appendix. T-statistics are in brackets. The superscript *** denotes significance at the 1% level, ** at the 5% level and * at the 10% level. Standard errors are clustered at the firm level. The standard errors clustered at the firm level for Columns (1) and (2) were calculated using OLS with the optimal λ. The sample period is 2005 to 2011.

Variables % Target Date Fund Int’l divers. bmk Trend Trend2 Cohort MN Experienced Ret

(1) Idiv 0.0688*** [7.77] 0.253* [1.80] 0.155 [0.91] 0.001 [0.21] 0.194*** [9.54] -0.554** [-2.37]

Return Chasing

(2) Idiv 0.0686*** [7.81] 0.201 [1.28] 0.080 [0.75] 0.005 [0.66] 0.167*** [13.53]

(3) Idiv 0.0687*** [7.60] 0.156 [0.72] 0.0616 [0.29] 0.00551 [0.78] 0.168*** [10.60]

(4) Idiv 0.0608*** [12.28] 0.294*** [2.78] -0.0881 [-0.65] 0.00642 [1.25] 0.156*** [14.19]

(5) Idiv 0.0690*** [9.24] 0.182 [1.02] -0.0592 [-0.34] 0.00854 [1.40] 0.161*** [12.13]

-0.243 [-0.08]

-1.260 [-0.81]

-0.441 [-0.18]

Flight to Safety

Constant Observations Adjusted R-squared Firm Fixed Effects Zip Code Fixed Effects

(7) Idiv 0.0591*** [11.76] 0.252*** [2.87] 0.00110 [0.01] 0.00452 [0.94] 0.158*** [15.40]

(8) Idiv 0.0681*** [9.42] 0.138 [0.80] 0.0501 [0.25] 0.00545 [0.84] 0.161*** [12.87]

-0.930 [-0.49]

0.607 [1.13]

-1.071 [-0.70]

-6.342 [-0.56] 17,082,302 0.053 N N

-10.63** [-1.97] 17,082,302 0.131 Y N

-4.513 [-0.46] 17,063,721 0.086 N Y

0.098 [1.21]

Relative returns

λ

(6) Idiv 0.0683*** [7.76] 0.151 [0.76] 0.114 [0.51] 0.00379 [0.54] 0.167*** [11.19]

3.99*** [4.11] -14.046* [-1.75] 17,426,447 0.054 N N

1.00 [0.21] -10.332 [-1.24] 17,426,477 0.054 N N

-6.581 [-0.53] 13,761,372 0.050 N N

-12.49* [-1.85] 13,761,372 0.128 Y N

-6.701 [-0.64] 13,747,036 0.086 N Y

65

Table 5 Income, Wealth and International Diversification This table reports the results for individual level regressions of international diversification on a quadratic time trend, birth year cohort and wealth variables (annual salary, 401(k) account value and the house value corresponding to the individual’s Zip code), all controlling for the percent invested in a target date fund and the international diversification benchmark. House values are either from Zillow (Columns (1) and (2), or from the Census (Columns (4) and (5)). All variables are defined in the Appendix. T-statistics are in brackets. The superscript *** denotes significance at the 1% level, ** at the 5% level and * at the 10% level. Standard errors are clustered at the firm level. The sample period is 2005 to 2011.

Variables % in Target Date Fund Intl Div. Benchmark Trend Trend2 Cohort ln(Annual Salary) ln(Annual Salary)2 ln(Account Value) ln(Account Value)2 ln(House Value Zillow)

(1) Idiv 0.0642*** [6.993] 0.378** [2.364] -0.0481 [-0.296] 0.00792 [1.313] 0.147*** [8.525] 0.138 [0.726] 0.191*** [4.727] 0.117 [0.949] -0.0511*** [-3.022] 0.697** [2.277]

(2) Idiv 0.0543*** [10.89] 0.317*** [3.522] 0.0589 [0.593] 0.00242 [0.561] 0.148*** [9.208] 0.185* [1.886] 0.186*** [8.601] 0.0551 [0.339] -0.0331** [-2.045] 0.653*** [4.837]

(3) Idiv 0.0653*** [8.200] 0.329*** [2.726] -0.119 [-0.859] 0.0100* [1.875] 0.140*** [10.27] 0.169 [1.292] 0.154*** [4.254] 0.0930 [0.814] -0.0470*** [-2.949]

(4) Idiv 0.0694*** [7.109] 0.335** [2.296] -0.0877 [-0.561] 0.00945 [1.627] 0.149*** [8.930] 0.0724 [0.333] 0.189*** [4.661] 0.108 [0.931] -0.0438** [-2.511]

(5) Idiv 0.0571*** [11.00] 0.300*** [3.756] 0.00590 [0.0600] 0.00413 [0.992] 0.147*** [9.796] 0.185** [2.035] 0.183*** [8.868] 0.0206 [0.137] -0.0277* [-1.696]

0.838*** [6.381] -26.10*** [-4.390]

ln(House Value Census) Constant

Observations Adjusted R-squared Firm Fixed Effects Zip Code Fixed Effects



-29.42*** [-2.684]

-25.33*** [-3.823]

-16.55** [-2.247]

0.975** [2.527] -30.04*** [-2.733]

10,621,481 0.047 N

10,621,481 0.120 Y

13,068,893 0.086 N

12,883,608 0.053 N

12,883,608 0.131 Y

N

N

Y

N

N

66

Table 6 International Diversification and Financial Advice Panel A – Summary Statistics This table reports the mean, median, std dev, 25th and 75th percentiles and number of observations for the individual level data stratified according to whether the individuals have signed the investor service agreement for online advice. The sample includes individuals with a positive equity allocation in their 401(k) portfolio. All variables are defined in the Appendix. The sample period is 2005 to 2011. Mean

Median

Std Dev

p25

p75

# Obs

17.335

16.667

15.784

3

25.714

15,571,691

16.834

0

34.586

0

1

15,571,691

63.668

63

11.745

55

73

15,571,691

45.471

46

11.735

37

54

15,571,691

56,420

46,116

47,388

32

70

11,792,124

58,735

21,184

107,211

5

68

15,548,551

243,629

186,651

189,469

125,421

299,386

12,697,864

239,545

186,923

164,297

128,535

297,968

15,332,097

21.011

20.000

16.363

9

29.670

1,293,508

7.739

0

21.516

0

0

1,293,508

63.187

63

10.902

55

71

1,293,508

45.855

46

10.842

38

54

1,293,508

73,095

64,766

49,153

45

92

1,110,295

124,977

71,811

161,535

23

166

1,288,957

242,834

199,792

170,824

135,700

292,611

1,095,710

247,129

207,414

153,304

143,092

304,964

1,274,765

17.617

16.981

15.859

4

26.087

16,865,199

16.137

0

33.850

0

0

16,865,199

63.631

63

11.683

55

72

16,865,199

45.500

46

11.669

37

54

16,865,199

57,855

47,498

47,772

33

73

12,902,419

63,806

23,386

113,675

6

75

16,837,508

243,566

187,780

188,056

126,176

298,762

13,793,574

240,127

188,785

163,492

129,561

298,625

16,606,862

Advice Dummy=0

International Diversification % Target Date Fund Cohort Age Annual Salary Total Account Value House Value (Zillow) House Value (Census) Advice Dummy=1

International Diversification % Target Date Fund Cohort Age Annual Salary Total Account Value House Value (Zillow) House Value (Census) Total

International Diversification % Target Date Fund Cohort Age Annual Salary Total Account Value House Value (Zillow) House Value (Census)



67

Panel B – Financial Advice, Its Recency, and Interactions with Individual Characteristics This panel replicates the regressions in Panel A of Table 5, controlling for whether the individual has signed up for online financial advice from Financial Engines and how recent she has logged onto the online advice website. All variables are defined in the Appendix. T-statistics are in brackets. The superscript *** denotes significance at the 1% level, ** at the 5% level and * at the 10% level. Standard errors are clustered at the firm level. The sample period is 2005 to 2011.

Variables % in Target Date Fund Intl Div. Benchmark Trend Trend2 Cohort Advice Dummy

(1) Idiv 0.0557*** [10.91] 0.321*** [3.630] 0.0213 [0.216] 0.00345 [0.811] 0.142*** [9.433] 4.273*** [10.29]

Non-Stale Advice Dummy

(2) Idiv 0.0557*** [10.88] 0.319*** [3.603] 0.0357 [0.352] 0.00313 [0.722] 0.142*** [9.454] 3.058*** [7.920] 2.246*** [5.320]

(3) Idiv 0.0668*** [8.375] 0.316*** [2.735] -0.126 [-0.889] 0.0101* [1.836] 0.134*** [10.40] 4.376*** [6.561]

(4) Idiv 0.0669*** [8.402] 0.313*** [2.705] -0.116 [-0.814] 0.00980* [1.763] 0.135*** [10.35] 2.690*** [5.486] 3.152*** [3.603]

Advice Dummy*Cohort ln(Annual Salary) ln(Annual Salary)2

0.123 [0.943] 0.148*** [4.099]

0.120 [0.915] 0.147*** [4.086]

0.0246 [0.167] -0.0465*** [-3.142]

0.0229 [0.158] -0.0451*** [-3.128]

0.0703 [0.630] -0.0622*** [-4.049]

0.0758 [0.675] -0.0607*** [-4.045]

0.700*** [5.477] -25.43*** [-3.925]

0.699*** [5.506] -25.43*** [-3.904]

-15.19** [-2.197]

-15.13** [-2.179]

10,621,481 0.126 Y N

10,621,481 0.126 Y N

13,068,893 0.093 N Y

13,068,893 0.094 N Y

10,621,481 0.126 Y N

Adv. Dmy*ln(Account Value)2

Constant

Observations Adjusted R-squared Firm Fixed Effects Zip Code Fixed Effects



13,068,893 0.093 N Y

0.143 [1.476] 0.181*** [8.318]

Adv. Dmy*ln(Account Value)

ln(House Value Zillow)

-0.0675*** [-4.532] 0.127 [0.942] 0.147*** [3.962] -0.120 [-0.429] 0.0125 [0.248] 0.0679 [0.619] -0.0586*** [-3.828] 0.115 [0.804] -0.0557** [-2.254]

0.143 [1.467] 0.182*** [8.309]

Adv. Dmy*ln(Annual Salary)2

ln(Account Value)2

(6) Idiv 0.0667*** [8.369] 0.316*** [2.744] -0.127 [-0.905] 0.0102* [1.849] 0.140*** [10.45] 9.501*** [7.312]

-0.0450*** [-3.536] 0.150 [1.504] 0.185*** [8.052] -0.0793 [-0.327] -0.0285 [-0.675] 0.0135 [0.0935] -0.0468*** [-3.115] 0.323** [2.500] -0.0356* [-1.813] 0.700*** [5.479] -25.76*** [-3.956]

Adv. Dmy*ln(Annual Salary)

ln(Account Value)

(5) Idiv 0.0554*** [10.84] 0.322*** [3.636] 0.0200 [0.203] 0.00349 [0.820] 0.146*** [9.391] 7.396*** [5.385]

68

-15.64** [-2.248]

Table 7 The Geography of International Diversification Panel A - Summary statistics Panel A presents the mean, median, standard deviation, 25th and 75th percentiles and number of observations for the Zip code level data. All variables are defined in the Appendix.

Variables Bachelor's Degree or Higher High School Degree College Degree Advanced Degree Financial Literacy Foreign Born Population Foreign Born Population - Latin America Foreign Born Population - Europe Foreign Born Population - Asia Distance to International Cities Distance to Tokyo Distance to London Distance to Mexico City Distance to Toronto Rural Long Distance Minutes State Exports/GDP State Openness GDP per capita GDP Growth 2000-2005 GDP Growth 2006-2011 House Value – Zillow House Value – Census



Mean 21.8 63.4 14.0 7.8 2.9 5.8 2.936 0.955 1.462 13,070 6,323 4,210 1,647 890 2.0 47 7.2 20.4 41,861 11.4 2.9 203,117 172,967

Median 17.4 65.6 11.9 5.4 2.9 2.2 0.421 0.295 0.182 12,801 6,515 4,143 1,655 705 1.0 46 6.6 17.8 40,451 11.3 2.6 156,350 125,900

Std Dev 16.0 14.2 9.8 8.4 0.1 9.2 6.622 2.077 3.830 790 624 596 451 647 1.2 7 3.2 9.2 10,525 5.4 6.2 166,527 145,372

p25 11.2 56.9 7.5 2.8 2.8 0.4 0.0 0.0 0.0 12,550 6,064 3,764 1,351 392 1.0 45 5.3 15.0 36,715 7.2 -0.2 110,148 84,900

p75 28.0 72.2 18.6 9.8 3.0 6.9 2.5 1.1 1.2 13,434 6,721 4,604 1,908 1,188 3.0 49 8.2 25.0 45,671 14.7 5.0 264,033 204,300

# Obs 32,746 32,746 32,746 32,746 42,107 32,751 32,751 32,751 32,751 41,631 41,631 41,631 41,631 41,631 41,982 42,107 42,107 42,107 42,107 42,107 42,107 12,446 31,921

69

Panel B - International Diversification Results The regressions in this table examine the Zip code fixed effects extracted from an individual level regression of international diversification on the percent invested in a target date fund, the international diversification benchmark, a quadratic time trend, birth year cohort, quadratic annual salary, quadratic account value advice variable and their interactions with salary and account values. Columns (2) and (4) include house values from the Zillow sample; Column (5) replicates Column (4) including house values from the 2010 Census, Columns (1), (3) and (6) do not include house values and are based on all the Zip codes in our sample. All variables are defined in the Appendix. T-statistics are in brackets. The superscript *** denotes significance at the 1% level, ** at the 5% level and * at the 10% level.

Variables Bachelor's or Higher

(1) Zip Code FE 0.0471*** [15.57]

(2) Zip Code FE 0.0473*** [15.96]

High School Degree Bachelor's Degree Advanced Degree Financial Literacy Foreign Born Population

4.051*** [11.22] 0.0290*** [5.489]

2.357*** [6.983] 0.0239*** [6.869]

Foreign Born - LatAm Foreign Born - Europe Foreign Born - Asia Distance to Intl Cities (,000 miles) Distance to Tokyo (,000 miles) Distance to London (,000 miles)



-0.124* [-1.845]

(3) Zip Code FE

(4) Zip Code FE

(5) Zip Code FE

(6) Zip Code FE

0.0318*** [5.417] 0.0634*** [8.901] 0.0633*** [7.199] 2.644*** [5.298]

0.0349*** [5.089] 0.0640*** [8.317] 0.0791*** [8.247] 4.387*** [9.882]

0.0345*** [5.432] 0.0639*** [7.627] 0.0772*** [7.424] 2.439*** [4.872]

0.0405*** [5.866] 0.0702*** [9.106] 0.0892*** [9.205] 4.202*** [9.044]

0.0339*** [4.415] 0.0985*** [3.765] 0.0465*** [3.613]

0.0462*** [7.575] 0.0966*** [5.116] 0.0186** [2.379]

0.0414*** [5.144] 0.0893*** [3.231] 0.0259* [1.945]

0.0463*** [7.562] 0.0952*** [5.020] 0.0133* [1.691]

-0.622*** [-3.103] -0.837*

0.833*** [5.503] 2.605***

-0.695*** [-3.418] -0.974**

0.686*** [4.517] 1.169***

0.000104 [0.00198]

70

Distance to Mexico (,000 miles) Distance to Toronto (,000 miles) Urban Large Rural Small Rural Long Distance Minutes State Exports/GDP

-0.374*** [-3.116] -0.336** [-2.286] -0.136 [-0.861] -0.0495*** [-4.418] 0.0947*** [7.130]

-0.234 [-1.290] -0.237 [-1.122] -0.497** [-2.058] 0.0372*** [4.896] 0.143*** [12.78]

[-1.911] -0.591** [-2.170] 0.289 [0.817] -0.394*** [-3.247] -0.344** [-2.332] -0.107 [-0.676] -0.0358*** [-2.760] 0.0892*** [6.239]

[7.312] 1.369*** [6.343] -1.851*** [-6.183] -0.327* [-1.796] -0.309 [-1.462] -0.522** [-2.166] 0.0350*** [4.068] 0.190*** [14.61]

[-2.200] -0.665** [-2.355] 0.393 [1.100] -0.430*** [-3.473] -0.342** [-2.326] -0.120 [-0.764] -0.0370*** [-2.860] 0.0883*** [6.185]

State Openness

GDP per Capita GDP Growth 2000-2005 GDP Growth 2006-2011

0.0531*** [10.42] -3.15e-05*** [-5.422] 5.179*** [6.065] 0.830 [1.613]

ln(House Value Zillow)

-4.50e05*** [-9.973] 2.319*** [2.849] 2.804*** [7.609] 0.174* [1.918]

-3.12e05*** [-5.363] 5.052*** [5.783] 0.420 [0.787]

-4.16e05*** [-9.174] 2.630*** [2.953] 3.534*** [8.745] 0.140 [1.396]

ln(House Value Census) Constant

Observations Adjusted R-squared



[3.527] 0.543*** [2.691] -0.484* [-1.767] -0.211 [-1.158] -0.278 [-1.313] -0.481** [-1.989] 0.0243*** [2.829]

-3.57e-05*** [-6.130] 5.111*** [5.865] 0.304 [0.571]

-24.97*** [-18.51]

-27.67*** [-21.99]

-17.40*** [-4.675]

-53.24*** [-15.94]

-0.0260 [-0.220] -15.45*** [-4.035]

28,525 0.020

12,297 0.085

28,525 0.022

12,297 0.092

28,136 0.022

-4.30e-05*** [-9.355] 1.122 [1.272] 4.105*** [9.869] -0.0924 [-0.924]

-42.26*** [-13.30] 12,297 0.084

71

Table 8 The Firm and International Diversification Panel A - Summary statistics on firm characteristics Panel A presents the mean, median, standard deviation, 25th and 75th percentiles and number of observations for the firm level data. For the private, foreign headquarter, foreign subsidiary dummies and the % of foreign subsidiaries variable, we substitute the median with the average of the 49th-51st percentiles, the 25th percentile with the average of the 24th- 26th percentiles, and the 75th percentile with the average of the 74th-76th percentiles. All variables are defined in the Appendix.

Variables Private Dummy Foreign Headquarter Dummy Foreign Subsidiary Dummy Foreign Subsidiaries (%) Industry Openness Firm Age # Employees Assets (USD mn) Leverage (%) Sales/Assets (%) Profitability (%) Investment Intensity (%) Fraction of Intnl Eq Funds (%) Expense Ratio of Intnl-Domestic (%) Turnover of Intnl/Domestic Eq Funds Alpha of Intnl- Alpha of Domestic Eq Funds (%) Fund Age of Intnl/Domestic Eq Funds High Fee Plan Dummy Total Plan Assets (USD mn)



Mean 0.62 0.16 0.56 28.6 24.0 69 18,623 38,693 30.6 106 2.84 4.18 21.47 .046 0.896 -0.004 0.932 0.481 456.93

Median 1.00 0.00 1.00 10.5 0.0 65 4,650 3,674 27.9 78.3 2.74 3.53 20.00 .0265 0.734 -0.004 0.854 0 332.79

Std Dev 0.48 0.37 0.50 33.5 35.9 45 48,093 200,300 20.6 120 9.67 3.28 7.57 .215 0.611 0.009 0.584 0.501 720.04

p25 0.00 0.00 0.00 0.0 0.0 28 1,732 1,319 16.9 40.5 0.89 1.86 16.67 -.008 0.493 -1.228 0.579 0 161.64

p75 1.00 0.00 1.00 69.1 44.6 102 14,730 25,243 39.3 127.6 5.91 5.85 25.56 .163 1.209 3.548 1.098 1 487.01

# Obs 290 290 289 289 264 268 265 156 126 152 156 125 296 296 294 296 291 296 296

72

Panel B - Firm characteristics and diversification The regressions in this table examine the firm fixed effects extracted from an individual level regression of international diversification on the percent invested in a target dated fund, the international diversification benchmark, a quadratic time trend, birth year cohort, advice dummy, and quadratic annual salary, quadratic account value, interacted with the advice dummy, and the house value corresponding to the individual’s Zip code. All variables are defined in the Appendix. Tstatistics are in brackets. The superscript *** denotes significance at the 1% level, ** at the 5% level and * at the 10% level.

Variables Private Public Parent Foreign Headquarters Dummy Foreign Subsidiaries Dummy ln(Firm Age) ln(# Employees) ln(Assets) Leverage Sales/Assets Profitability Investment Intensity



(1) Firm FE 3.922*** [6.017] -2.239** [-2.326] 1.341 [1.332]

(2) Firm FE 3.911* [1.808] 0.112 [0.0315] 3.981 [1.059] 3.591** [2.241] 0.538 [0.900] 0.620 [0.985] -0.397 [-0.678] -0.00476 [-0.162] -0.0106 [-0.738] -0.172** [-2.241] -0.0196 [-0.110]

(3) Firm FE 5.347** [2.239] -2.395 [-0.637] -4.683 [-0.859] 3.936** [2.194] 0.673 [1.016] 0.899 [0.975] -0.580 [-0.694] -0.00507 [-0.146] -0.00400 [-0.265] -0.138* [-1.678] -0.215 [-0.864]

(4) Firm FE 3.918* [1.797] 0.126 [0.0349] 3.966 [1.046] 3.571** [2.142] 0.542 [0.891] 0.618 [0.976] -0.390 [-0.645] -0.00452 [-0.151] -0.0106 [-0.726] -0.172** [-2.226] -0.0206 [-0.114]

(5) Firm FE

0.322 [0.544] 0.230 [0.406] -0.134 [-0.255] 0.00470 [0.175] 0.00220 [0.161] -0.157** [-2.120] -0.110 [-0.708]

(6) Firm FE 3.857* [1.712] 0.361 [0.0981] 2.949 [0.761] 3.001* [1.780] 0.698 [1.081] 0.644 [0.937] -0.530 [-0.846] 0.0105 [0.331] -0.0111 [-0.717] -0.185** [-2.360] -0.0809 [-0.418]

73

Industry Openness

0.000625 [0.0464]

0.00219 [0.172]

-34.05*** [-6.989] 109 0.165

-29.04*** [-6.437] 113 0.057

0.00334 [0.231] 5.068 [0.614] -5.427* -0.120 -0.145 70.29 [1.328] 0.488 0.221 -0.120 3.309*** [2.801] 2.31e-07 [0.202] -25.73*** [-4.783] 103 0.289

N

N

N

Fraction of Intnl Eq Funds Expense Ratio of Intnl- Domestic Eq Funds Turnover of Intnl/Domestic Eq Funds Alpha of Intnl-Domestic Eq Funds Fund Age of Intnl/Domestic Eq Funds High Fee Plan Dummy





Total Plan Assets

-27.40***

-33.98***

[-58.11] 286 0.120

[-7.453] 109 0.165

-31.68*** [-4.093] 109 0.345

N

N

Y

Constant Observations R-squared Industry Fixed Effects





74

Table 9 Panel A - Controlling for the Quarter the Worker Joined the Firm, the Firm’s Identity and the Quarter-Year of Observation This table reports the results of including fixed effects based on the quarter the worker joined the firm, the firm’s identity and the quarter-year of observation in the regressions in Column (4) of Table 3, Columns (2) and (5) of Table 5 and of Table 6. All variables are defined in the Appendix. T-statistics are in brackets. The superscript *** denotes significance at the 1% level, ** at the 5% level and * at the 10% level. Standard errors are clustered at the firm level. The sample period is 2005 to 2011.

Variables % in Target Date Fund Intl Div. Benchmark Trend Trend2 Cohort

(1) Idiv 0.0324*** [4.645] 0.257*** [2.624] 0.156 [1.348] -0.00234 [-0.482] 0.0478*** [5.608]

(2) Idiv 0.0304*** [4.841] 0.312*** [2.822] 0.207* [1.795] -0.00583 [-1.103] 0.0566*** [8.137]

(3) Idiv 0.0322*** [5.215] 0.275*** [2.639] 0.197* [1.685] -0.00494 [-0.944] 0.0572*** [7.756]

Advice Dummy Non-Stale Advice Dummy

(4) Idiv 0.0321*** [5.235] 0.315*** [2.896] 0.186 [1.545] -0.00523 [-0.964] 0.0543*** [8.065] 2.415*** [11.16] 1.638*** [5.292]

Advice Dummy*Cohort ln(Annual Salary) ln(Annual Salary)2

0.283** [2.450] 0.0726*** [3.506]

0.284** [2.454] 0.0774*** [3.699]

0.257** [2.239] 0.0697*** [3.409]

0.248*** [4.042] 0.0603*** [3.489]

0.213*** [3.411] 0.0642*** [3.625]

0.233*** [3.726] 0.0503*** [2.964]

Adv. Dmy*ln(Annual Salary) Adv. Dmy*ln(Annual Salary)2 ln(Account Value) ln(Account Value)2 Adv. Dmy*ln(Account Value) Adv. Dmy*ln(Account Value)2 ln(House Value Zillow) ln(House Value Census)



0.238** [2.072]

0.260** [2.247]

(5) Idiv 0.0315*** [5.084] 0.315*** [2.888] 0.172 [1.484] -0.00493 [-0.927] 0.0569*** [7.827] 4.593*** [4.306]

-0.0264** [-2.610] 0.269** [2.351] 0.0706*** [3.359] -0.295 [-1.010] 0.0171 [0.505] 0.226*** [3.492] 0.0428*** [2.867] 0.271* [1.792] 0.0263 [1.340] 0.257** [2.224]

0.348*** [3.464]

75

Constant Observations Adjusted R-squared Quarter Joined *Quarter-Year *Firm Fixed Effects

-2.664 [-0.436] 9,170,900 0.136

-12.78* [-1.699] 6,040,811 0.134

-12.02* [-1.725] 7,038,080 0.136

-13.03* [-1.748] 6,040,811 0.138

-12.96* [-1.749] 6,040,811 0.138

Y

Y

Y

Y

Y

Panel B - Controlling for the International Funds offered and their Quality The Table repeats the regressions in Column (4) of Table 3, Columns (2) and (5) of Table 5 and of Table 6, controlling directly for the number and quality of the international funds offered by the plan in a more recent subsample. All variables are defined in the Appendix. T-statistics are in brackets. The superscript *** denotes significance at the 1% level, ** at the 5% level and * at the 10% level. Standard errors are clustered at the firm level. The sample period is 2005 to 2011. The omitted controls are as follows: {A: TDF, benchmark, Trend, Cohort}; {B: Salary, Account value, House value Zillow}; {B’: Salary, Account value, House value Census}; {C: Advice and Non-Stale Advice}; {D: Advice Dummy and Interactions}.

Variables Controls for TDF, benchmark, Trend, Cohort, Salary, Account value, House value, Advice dummy and its interactions. Fraction of Intnl Eq Funds

Expense Ratio of Intnl - Domestic Turnover of Intnl/Dom Eq Funds Alpha Intnl- Alpha Dom Eq Funds Fund Age of Intnl/Dom Eq Funds High Fee Plan Dummy Total Plan Assets Constant Observations Adjusted R-squared



(1)

(2)

(3)

(4)

(5)

Idiv

Idiv

Idiv

Idiv

Idiv

A

A+B

A+B’

A+B+C

A+B+D

0.177*** [2.850] -4.305** [-1.998] -0.212 [-0.268] 8.889 [0.134] -0.178 [-0.360] 3.009** [2.541] 5.98e-07* [1.815] -11.71 [-1.452] 16,299,381

0.204*** [3.403] -3.950** [-2.068] -0.135 [-0.158] 11.77 [0.154] -0.192 [-0.401] 2.408* [1.820] 8.18e-08 [0.135] -27.61*** [-2.854] 9,983,354

0.202*** [3.362] -4.240** [-2.227] -0.112 [-0.133] 30.00 [0.425] -0.222 [-0.470] 2.583** [2.062] 2.02e-07 [0.507] -28.20*** [-2.984] 12,163,670

0.198*** [3.505] -3.929** [-2.059] -0.158 [-0.182] 6.374 [0.0822] -0.200 [-0.429] 2.497* [1.956] 2.60e-07 [0.476] -26.07*** [-2.813] 9,983,354

0.199*** [3.517] -3.877** [-2.032] -0.199 [-0.227] 6.258 [0.0806] -0.187 [-0.400] 2.471* [1.921] 2.36e-07 [0.424] -26.82*** [-2.873] 9,983,354

0.066

0.058

0.065

0.065

0.065

76

Table 10 Subsamples This table reports the results for individual level regressions of international diversification on the percent invested in a target dated fund, the international diversification benchmark, a quadratic time trend and birth year cohort for different subsamples. Panel B reports the results of these same regressions, controlling for salary, wealth and access to advice effects. Column (2) excludes observations with (a) tenure 0-3, age>35, (b) tenure 4-5, age>40, (c) tenure 6-10, age>45, (d) tenure 11-15, age>50, (e) tenure 16-20, age>55, (f) those with missing tenure. Column (3) excludes observations with salaries>=100,000 and account balances>=200,000, along with those that have missing information for either variable. Column (4) is a combination of the exclusion rules specified in Columns (2) and (3). Column (5) excludes observations with bond allocations over 50% and Column (6) uses international stocks as the dependent variable. All variables are defined in the Appendix. T-statistics are in brackets. The superscript *** denotes significance at the 1% level, ** at the 5% level and * at the 10% level. Standard errors are clustered at the firm level. The sample period is 2005 to 2011.

Variables % in Target Date Fund Intl Div. Benchmark Trend Trend2 Cohort Constant Observations Adjusted R-squared Firm Fixed Effects

Subsample



(1) Idiv 0.0590*** [11.79] 0.214*** [2.759] 0.0395 [0.348] 0.00331 [0.753] 0.158*** [15.41] -8.450* [-1.736] 17,082,302 0.131 Y

(2) Idiv 0.0481*** [7.683] 0.188 [1.632] 0.133 [0.863] -0.000275 [-0.0442] 0.158*** [10.39] -5.691 [-0.819] 11,508,410 0.115 Y

(3) Idiv 0.0539*** [11.14] 0.245*** [3.373] -0.00786 [-0.0779] 0.00532 [1.298] 0.152*** [12.20] -9.492* [-1.967] 14,118,739 0.125 Y

(4) Idiv 0.0445*** [7.125] 0.243** [2.610] 0.0287 [0.243] 0.00419 [0.815] 0.147*** [8.551] -7.922 [-1.290] 9,847,022 0.109 Y

(5) Idiv 0.0560*** [11.15] 0.209*** [2.737] 0.0815 [0.694] 0.00178 [0.397] 0.161*** [17.27] -8.439* [-1.762] 15,940,134 0.138 Y

(6) Intl Stock 0.0709*** [20.15] 0.278*** [4.499] -0.318*** [-3.859] 0.0128*** [3.933] 0.211*** [16.28] -18.90*** [-4.809] 19,017,474 0.142 Y

Whole Sample

Age-Tenure Screen

Salary-Acct Value Screen

Age/Tenure & Salary/Acct

Exclude High Bond Alloc.

Intl Stock as Dep Var

77

Online Appendix Appendix 1: Firm Characteristics Panel A presents the mean, median, standard deviation and 25th and 75th percentiles for firms in our sample between 2005 and 2011. The 25th percentile is an average of the 24th, 25th and 26th percentiles, the median is the average of the 49th, 50th, and 51st percentiles and the 75th percentile is the average of the 74th, 75th, and 76th percentile. Panels B and C present these same statistics for all firms in Compustat and the S&P 500 between 2005 and 2011, respectively. Note that firm age in these two cases is calculated as number of years in Compustat.

Panel A - Sample Firms Variables

Mean

Median

Std Dev

p25

p75

Assets (USD mn)

43,139

4,096

223,921

1,268

25,007

Debt (USD mn)

25,214

1,770

187,098

573

6,084

284

129

3,483

6

696

9,484

2,919

17,361

1,150

9,457

Net Income (USD mn) Sales (USD mn) Capex (USD mn)

725

149

1,202

48

873

Leverage (%)

30.64

27.91

20.61

17

40

Sales/Assets (%)

105.96

78.26

119.92

39

128

Profitability (%)

2.84

2.74

9.67

1

6

ROA

4.18 3.02%

3.53 2.71%

3.28 9.40%

2 1%

6 6%

ROE

0.60%

3.93%

21.96%

1%

7%

Annual Return (%)

10.95%

7.37%

27.64%

1%

17%

18,623

4,650

48,093

1,722

14,730

69

65

45

28

103

456.93

332.79

720.04

161.64

487.01

Investment Intensity (%)

Number of Employees Firm Age (years) Plan size (USD mn)





78

Panel B - Compustat Firms Variables

Mean

Median

Std Dev

p25

p75

Assets (USD mn)

14,550

401

121,056

51

2,054

Debt (USD mn)

4,063

42

44,738

1

457

180

3

1,516

-5

46

Sales (USD mn)

2,790

124

13,590

15

893

Capex (USD mn)

204

5

1,202

0

44

Leverage (%)

111.43

15.55

3280.24

1.86

35.58

Sales/Assets (%)

106.36

53.69

5353.48

11.15

111.22

Profitability (%)

-424.14

0.99

54130.84

-9.62

5.83

Investment Intensity (%)

6.10

2.35

50.08

0.48

6.29

ROA

-4.24

0.01

541.31

-0.10

0.06

ROE

-8.4

3.9

45.4

-16.1

5.9

Annual Return (%)

5.65

0.14

79.28

-25.31

22.78

Number of Employees

8,058

475

36,694

93

3,046

14

9

14

5

20

Variables

Mean

Median

Std Dev

p25

p75

Assets (USD mn)

76,341

14,604

236,713

5,839

42,443

Debt (USD mn)

22,956

3,196

93,858

1,041

8,342

Net Income (USD mn)

1,301

576

4,033

246

1,340

Sales (USD mn)

17,891

7,785

33,907

3,557

16,411

Net Income (USD mn)

Firm Age (years in Compustat)

Panel C - S&P 500 Firms

Capex (USD mn)

1,001

271

2,322

99

931

Leverage (%)

23.93

21.34

17.73

10.53

34.16

Sales/Assets (%)

83.75

67.31

72.35

36.51

107.12

Profitability (%)

5.84

5.41

8.00

2.22

9.57

Investment Intensity (%)

4.17

3.05

4.54

1.28

5.57

ROA

5.84

5.41

8.00

2.22

9.57

ROE

7.90

7.37

39.16

3.10

12.34

Annual Return (%)

6.45

5.32

39.96

-14.65

24.49

44,355

18,028

101,793

6,900

43,836

35

35

19

14

45

Number of Employees Firm Age (years in Compustat)



79

Panel D Sample Firms – Private versus Public Panel D presents the mean, median, standard deviation, 25th and 75th percentiles and number of firm-year observations in our sample between 2005 and 2011. The 25th percentile is an average of the 24th, 25th and 26th percentiles, the median is the average of the 49th, 50th, and 51st percentiles and the 75th percentile is the average of the 74th, 75th, and 76th percentile. Summary statistics are decomposed into private firms and public firms. There are 178 private firms, 108 public firms and 4 firms who switch from public to private or private to public in the sample.

Public Firms Variable Assets (USD mn) Debt (USD mn) Net Income (USD mn) Sales (USD mn) Capex (USD mn) Leverage (%) Sales/Assets (%) Profitability (%) Investment Intensity (%) ROA ROE Annual Return (%) Number of Employees Firm Age (years)

Mean 54,858 28,917 421 12,453 750 28.95 81.22 2.69 4.20 2.85% 2.38% 10.51% 35,200 80

Median 265,475 201,979 4,170 19,712 1,229 19.73 52.76 7.74 3.25 7.30% 15.15% 27.91% 70,626 47

Std Dev 6,879 1,836 227 4,106 166 26.66 70.15 2.96 3.52 2.88% 4.32% 7.22% 12,061 81

p25 2,483 608 54 1,742 53 16.83 38.93 1.01 1.95 1.16% 1.36% 1.41% 4,648 40

Std Dev 1,034 1,811 6 883 54 45.65 108.07 2.24 3.70 1.97% -0.65% 9.48% 2,556 62

p25 169 410 -12 171 34 24.32 41.70 -0.67 1.94 -0.68% -6.48% -8.69% 783 26

p75 32,060 7,084 1,114 14,065 907 37.79 113.74 6.27 5.91 6.23% 9.08% 16.64% 30,658 120



Private Firms Variable Assets (USD mn) Debt (USD mn) Net Income (USD mn) Sales (USD mn) Capex (USD mn) Leverage (%) Sales/Assets (%) Profitability (%) Investment Intensity (%) ROA ROE Annual Return (%) Number of Employees Firm Age (years)

Mean 18,112 3,672 -32 2,274 724 43.99 173.22 3.38 4.60 3.65% -13.90% 8.99% 7,279 63

Median 64,988 4,145 549 4,060 1,136 23.38 203.11 13.59 3.67 13.48% 49.60% 17.44% 13,255 42

80

p75 4,314 5,729 117 2,248 1,208 57.91 254.94 5.40 6.93 5.19% 3.86% 26.17% 6,543 94

Appendix 2: Employee Characteristics

Panel A Employee Characteristics across Firms Panel A presents the mean, median, standard deviation and 25th and 75th percentiles for all individuals in the sample between 2005 and 2011 (the data include both stock market participants and non-stock market participants). Variables Salary

Mean 46,205

Median 39,687

Std Dev 48.014

p25 18,879

p75 63,890

Total Account Value

62,798

22,255

113,850

5,332

73,334

Contribution Rate

5.89%

5.00%

6.16%

0%

8.00%

Tenure

10.55

7.25

10.64

2.02

16.22

46

46

12

37

54

1963

1963

12

1955

1972

Age Cohort

Panel B Current Population Survey (CPS) Panel B presents the mean, median, standard deviation and 25th and 75th percentiles for individual statistics in the Current Population Survey between 2006 and 2011. In order to extract tenure data, we use the January CPS Displaced Worker, Employee Tenure and Occupational Mobility Supplement for years 2006, 2008, and 2010, while 2007, 2009, and 2011 data come from the January CPS. The summary statistics reported in this table are the average of the annual statistics. Variables

Mean

Median

Std Dev

p25

p75

Salary

45,437

37,175

30,045

19,432

79,980

Tenure

7.7

5.0

8.2

1.42

19.6

Age

41

42

12

28

48

Panel C - Summary Statistics for Managed Accounts This table reports the mean, median, std dev, 25th and 75th percentiles and number of observations for the individual level data. The sample includes individuals with a managed account. All variables are defined in the Appendix. The sample period is 2005 to 2011. Variable

Mean

Median

Std Dev

p25

p75

# Obs

Cohort

1962

1961

11

1955

1972

1,611,453

Age

46

47

11

38

55

1,611,453

Annual Salary

56,160

47,625

42,147

27,040

60,807

1,363,806

Total Account Value

59,639

27,735

91,565

8,636

98,662

1,611,552

House Value (Census)

234,266

178,300

159,756

149,000

366,500

1,587,840

Tenure

8.1

3.7

9.2

4.4

20.6

1,476,011

Contribution Rate (%)

7

6

6

3

9.78

1,363,806

81

Appendix 3 Panel A - Controlling for the International Funds offered and their Quality The Table repeats the regressions in Column (4) of Table 3, Columns (2) and (5) of Table 5 and of Table 6, controlling directly for the number and quality of the international funds offered by the plan in a more recent subsample. All variables are defined in the Appendix. T-statistics are in brackets. The superscript *** denotes significance at the 1% level, ** at the 5% level and * at the 10% level. Standard errors are clustered at the firm level. The sample period is 2005 to 2011.

Variables % in Target Date Fund

Div. Benchmark Intl Trend Trend2 Cohort

(1)

(2)

(3)

(4)

(5)

Idiv

Idiv

Idiv

Idiv

Idiv

0.0630*** [8.213] 0.163 [1.240] 0.0892 [0.536] 0.00398 [0.713] 0.165*** [12.57]

0.0581*** [7.571] 0.274** [1.985] -0.0408 [-0.248] 0.00795 [1.385] 0.143*** [8.414]

0.0637*** [7.242] 0.255** [2.023] -0.0968 [-0.612] 0.00980* [1.755] 0.144*** [8.815]

0.0595*** [7.695] 0.258* [1.897] -0.0611 [-0.383] 0.00855 [1.503] 0.138*** [8.734] 2.608*** [3.362] 3.182** [2.578]

0.0594*** [7.700] 0.262* [1.925] -0.0695 [-0.440] 0.00885 [1.565] 0.142*** [8.832] 10.37*** [5.815]

Advice Dummy Non-Stale Advice Dummy Advice Dummy*Cohort ln(Annual Salary) ln(Annual Salary)2

0.0439 [0.161]

-0.0242 [-0.0841]

-0.00239 [-0.00872]

0.220*** [4.551]

0.219*** [4.847]

0.215*** [4.441]

0.0934 [0.720]

0.0796 [0.663]

0.0702 [0.565]

-0.0664*** [-3.717]

-0.0572*** [-3.182]

-0.0778*** [-4.536]

Adv. Dmy*ln(Annual Salary) Adv. Dmy*ln(Annual Salary)2 ln(Account Value) ln(Account Value)2 Adv. Dmy*ln(Account Value) Adv. Dmy*ln(Account Value)2 ln(House Value Zillow)

0.695** [2.382]

ln(House Value Census) Fraction of Intnl Eq Funds

0.177*** [2.850]

0.204*** [3.403] 82

0.872*** [2.660] 0.202*** [3.362]

-0.0654*** [-3.185] 0.00971 [0.0350] 0.216*** [4.324] -0.513 [-1.035] 0.0576 [0.546] 0.0533 [0.437] -0.0727*** [-4.317] 0.258* [1.713]

0.697** [2.448]

-0.0951*** [-3.085] 0.709** [2.472]

0.198*** [3.505]

0.199*** [3.517]

Expense Ratio of Intnl - Domestic

Turnover of Intnl - Dom Eq Funds Alpha Intnl- Alpha Dom Eq Funds Fund Age of Intnl/Dom Eq Funds High Fee Plan Dummy Total Plan Assets Constant Observations Adjusted R-squared

-4.305** [-1.998] -0.212 [-0.268] 8.889 [0.134] -0.178 [-0.360] 3.009** [2.541] 5.98e-07* [1.815] -11.71 [-1.452] 16,299,381

-3.950** [-2.068] -0.135 [-0.158] 11.77 [0.154] -0.192 [-0.401] 2.408* [1.820] 8.18e-08 [0.135] -27.61*** [-2.854] 9,983,354

-4.240** [-2.227] -0.112 [-0.133] 30.00 [0.425] -0.222 [-0.470] 2.583** [2.062] 2.02e-07 [0.507] -28.20*** [-2.984] 12,163,670

-3.929** [-2.059] -0.158 [-0.182] 6.374 [0.0822] -0.200 [-0.429] 2.497* [1.956] 2.60e-07 [0.476] -26.07*** [-2.813] 9,983,354

-3.877** [-2.032] -0.199 [-0.227] 6.258 [0.0806] -0.187 [-0.400] 2.471* [1.921] 2.36e-07 [0.424] -26.82*** [-2.873] 9,983,354

0.066

0.058

0.065

0.065

0.065

83

Appendix 4: Subsamples – Income, Wealth, Access to Advice and International Diversification Panel A reports the results for individual level regressions of international diversification on the percent invested in a target date fund, the intnl div benchmark, a quadratic time trend and birth year cohort for different subsamples, salary and wealth on various subsamples. Panel B adds access to online advice. See Table 10 for more details on this table.

Variables % in Target Date Fund Intl Div. Benchmark Trend Trend2 Cohort ln(Annual Salary) ln(Annual Salary)2 ln(Account Value) ln(Account Value)2 ln(House Value Zillow) Constant Observations Adjusted R-squared Firm Fixed Effects Subsample

(1) Idiv 0.0543*** [10.89] 0.317*** [3.522] 0.0589 [0.593] 0.00242 [0.561] 0.148*** [9.208] 0.185* [1.886] 0.186*** [8.601] 0.0551 [0.339] -0.0331** [-2.045] 0.653*** [4.837] -25.33*** [-3.823] 10,621,481 0.120 Y

(2) Idiv 0.0417*** [6.258] 0.322*** [2.655] 0.0588 [0.481] 0.00221 [0.382] 0.121*** [13.76] 0.292* [1.780] 0.149*** [4.646] -0.170 [-1.577] -0.0416** [-2.169] 0.684*** [4.835] -21.88** [-2.487] 6,040,610 0.094 Y

(3) Idiv 0.0538*** [10.82] 0.297*** [3.473] 0.0504 [0.507] 0.00306 [0.718] 0.150*** [9.158] 0.182* [1.845] 0.187*** [7.879] 0.0660 [0.412] -0.0403** [-2.262] 0.653*** [4.545] -24.07*** [-3.755] 10,216,034 0.122 Y

(4) Idiv 0.0412*** [6.239] 0.295** [2.598] 0.0486 [0.400] 0.00310 [0.549] 0.122*** [13.84] 0.293* [1.807] 0.148*** [4.317] -0.156 [-1.490] -0.0498** [-2.305] 0.689*** [4.820] -20.29** [-2.409] 5,813,961 0.096 Y

(5) Idiv 0.0509*** [10.01] 0.312*** [3.450] 0.109 [0.981] 0.000616 [0.132] 0.148*** [10.49] 0.187* [1.928] 0.181*** [8.281] 0.0697 [0.381] -0.0413** [-2.499] 0.643*** [4.562] -24.80*** [-3.728] 9,898,960 0.126 Y

(6) Int. Stock 0.0743*** [18.99] 0.347*** [4.308] -0.321*** [-3.932] 0.0125*** [3.514] 0.224*** [12.11] 0.222* [1.801] 0.165*** [7.944] 0.252** [2.306] -0.00535 [-0.460] 0.637*** [5.131] -35.56*** [-5.882] 11,642,469 0.138 Y

Whole Sample

Age-Tenure Screen

Salary-Acct Value Screen

Age/Tenure & Salary/Acct

Exclude High Bond Alloc.

Intl Stock as Dep Var

84

Panel B

Variables % in Target Date Fund Intl Div. Benchmark Trend Trend2 Cohort Advice Dummy Advice Dummy*Cohort ln(Annual Salary) ln(Annual Salary)2 Adv. Dmy*ln(Annual Salary) Adv. Dmy*ln(Annual Salary)2 ln(Account Value) ln(Account Value)2 Adv. Dmy*ln(Account Value) Adv. Dmy*ln(Account Value)2 ln(House Value Zillow) Constant Observations Adjusted R-squared Firm Fixed Effects

Subsample

(1) Idiv 0.0554*** [10.84] 0.322*** [3.636] 0.0200 [0.203] 0.00349 [0.820] 0.146*** [9.391] 7.396*** [5.385] -0.0450*** [-3.536] 0.150 [1.504] 0.185*** [8.052] -0.0793 [-0.327] -0.0285 [-0.675] 0.0135 [0.0935] -0.0468*** [-3.115] 0.323** [2.500] -0.0356* [-1.813] 0.700*** [5.479] -25.76*** [-3.956] 10,621,481 0.126 Y

(2) Idiv 0.0427*** [6.480] 0.326*** [2.712] 0.0225 [0.180] 0.00317 [0.540] 0.122*** [13.74] 5.783*** [3.968] -0.0399** [-2.348] 0.289* [1.822] 0.146*** [4.418] -0.535** [-2.183] 0.0569 [1.178] -0.198* [-1.853] -0.0562*** [-2.691] 0.530** [2.523] -0.0355 [-1.394] 0.703*** [4.745] -22.25** [-2.493] 7,425,729 0.108 Y

(3) Idiv 0.0550*** [10.77] 0.301*** [3.585] 0.0121 [0.122] 0.00412 [0.980] 0.147*** [9.331] 7.116*** [4.895] -0.0422*** [-3.396] 0.149 [1.484] 0.185*** [7.348] -0.0843 [-0.326] -0.0209 [-0.506] 0.0213 [0.149] -0.0528*** [-3.145] 0.355*** [3.270] -0.0413 [-1.527] 0.697*** [5.153] -24.43*** [-3.888] 10,621,481 0.126 Y

(4) Idiv 0.0422*** [6.467] 0.299*** [2.653] 0.0125 [0.101] 0.00407 [0.711] 0.123*** [13.67] 5.653*** [4.023] -0.0388** [-2.444] 0.292* [1.871] 0.143*** [4.091] -0.584** [-2.486] 0.0717* [1.679] -0.189* [-1.809] -0.0627*** [-2.719] 0.597*** [3.927] -0.0492 [-1.282] 0.706*** [4.743] -20.61** [-2.414] 7,425,729 0.108 Y

(5) Idiv 0.0519*** [10.06] 0.316*** [3.550] 0.0687 [0.627] 0.00172 [0.374] 0.147*** [10.65] 8.194*** [6.070] -0.0563*** [-4.751] 0.157 [1.609] 0.181*** [7.762] -0.139 [-0.553] -0.0329 [-0.780] 0.0272 [0.166] -0.0558*** [-3.706] 0.351*** [2.640] -0.0324* [-1.673] 0.690*** [5.147] -25.27*** [-3.867] 9,898,960 0.132 Y

(6) Intl Stock 0.0754*** [19.17] 0.351*** [4.38] -0.349*** [-4.24] 0.0132*** [3.73] 0.218*** [12.15] 2.961*** [2.92] 0.0118 [0.86] 0.193* [1.65] 0.161*** [7.44] -0.144 [-0.66] 0.0168 [0.55] 0.208** [2.11] -0.00938 [-0.85] 0.322*** [3.17] -0.0711*** [-3.92] 0.670*** [5.66] -35.57*** [-5.92] 11,642,469 0.143 Y

Whole Sample

AgeTenure Screen

SalaryAcct Value Screen

Age/Tenure & Salary/Acct

Exclude High Bond Alloc.

Intl Stock as Dep Var.

85

Appendix 5 – Tobit Regressions The table repeats the regressions in Column (4) of Table 3, Columns (2) and (5) of Table 5 and of Table 6 using a Tobit specification. All variables are defined in the Appendix. T-statistics are in brackets. The superscript *** denotes significance at the 1% level, ** at the 5% level and * at the 10% level. Standard errors are clustered at the firm level. The sample period is 2005 to 2011.

Variables % in Target Date Fund Intl Div. Benchmark Trend Trend2 Cohort

(1) Idiv 0.0885*** [8.190] 0.184 [0.988] 0.224 [1.032] 0.00162 [0.242] 0.229*** [10.01]

(2) Idiv 0.0833*** [7.544] 0.402** [2.180] 0.0573 [0.282] 0.00526 [0.753] 0.197*** [7.528]

(3) Idiv 0.0898*** [7.604] 0.356** [2.108] 0.0202 [0.104] 0.00684 [1.015] 0.202*** [7.719]

Advice Dummy Non-Stale Advice Dummy

(4) Idiv 0.0851*** [7.610] 0.371** [2.111] 0.0766 [0.378] 0.00454 [0.630] 0.188*** [7.467] 3.217*** [4.769] 3.600*** [3.245]

Advice Dummy*Cohort ln(Annual Salary) ln(Annual Salary)2

0.858** [2.161]

-0.0990*** [-3.859] 0.0877 [0.412] 0.212*** [4.037] 0.0102 [0.0270] 0.0234 [0.343] 0.195 [1.171] -0.0642*** [-2.909] 0.169 [0.770] -0.0921*** [-2.589] 0.875** [2.188]

-37.13*** [-3.170] 10,621,481

-38.03*** [-3.215] 10,621,481

0.148 [0.692] 0.222*** [4.447]

0.0687 [0.280] 0.218*** [4.369]

0.0835 [0.400] 0.213*** [4.248]

0.231 [1.366] -0.0497** [-2.163]

0.228 [1.412] -0.0399* [-1.710]

0.206 [1.213] -0.0684*** [-3.139]

Advice Dummy*ln(Annual Salary) Advice Dummy*ln(Annual Salary)2 ln(Account Value) ln(Account Value)2 Advice Dummy*ln(Account Value) Advice Dummy*ln(Account Value)2 ln(House Value Zillow)

0.836** [2.098]

ln(House Value Census) Constant Observations

-15.69 [-1.537] 17,082,302

-39.43*** [-3.144] 10,621,481 86



1.266** [2.441] -41.97*** [-3.225] 12,883,608

(5) Idiv 0.0849*** [7.597] 0.374** [2.140] 0.0673 [0.334] 0.00492 [0.687] 0.196*** [7.420] 12.01*** [5.219]

Appendix 6: Variable Description A 401(k) plan is a defined contribution retirement savings plan offered by many U.S. firms to their employees (401(k) refers to the subsection of the Internal Revenue Code which defines the plans). Employee contributions are made as deductions from their paychecks and are placed in an individual account for each employee within the plan. The firm typically provides a range of investment options from which each employee can chose. Savings in these accounts receive a variety of different preferential tax treatments and may also receive matching contributions from the firm. Individual Level Variables International Diversification (idiv)

Description Allocation to international equities over allocation to all equities. The total equity allocation is defined as the combination of investments in Large Cap Stocks, Small and Mid-Cap Stocks, Individual Stocks, Company Stock and International Stocks. This series is individual specific. Source: Financial Engines.

Cohort

The cohort variable is defined as the individual's birth year minus 1900. The cohort is set to 1993 if the individual is born after 1990 and to 1940 if the individual is born before 1945. This data is individual specific. Source: Financial Engines.

Age

Age is defined as the difference between the observation date and the individual's birth date. Source: Financial Engines.

Total Account Value (log)

Total account values represent the balance in the 401(k) account. This value is first deflated to 2005 prices using the Consumer Price Index for All Urban Consumers and then the natural logarithm is taken. Source: Financial Engines and U.S. Department of Labor: Bureau of Labor Statistics.

House Value - Zillow (log)

The natural logarithm of house values deflated to 2005 prices using the Consumer Price Index for All Urban Consumers. We match the Zillow average house value in a Zip code to each individual based on the Zip code they live in according to Financial Engines. Source: Zillow, U.S. Department of Labor: Bureau of Labor Statistics, Financial Engines.

House Value - Census (log)

The natural logarithm of median house values in dollars at the Zip code level. This variable is matched to the individual data using the Zip code where the user lives. Source: U.S. Census Bureau, 2008-2012 American Community Survey - Table B25077: Median Housing Value of Owner-Occupied Housing Units (Dollars). 87



Annual Salary (log)

Annual Salary represents the dollar amount an individual is paid by the company. The dollar amount is first deflated to 2005 prices using the Consumer Price Index for All Urban Consumers and then the natural logarithm is taken. Source: Financial Engines and U.S. Department of Labor: Bureau of Labor Statistics.

% Target Date Fund

Amount allocated to target dated funds as a percentage of the individual's total account value. This data is individual specific. Source: Financial Engines.

International Diversification Benchmark

The ratio of international market cap (MSCI Market Cap All Countries ex-US) to the sum of international and domestic market cap (MSCI Market Cap All Countries). We obtain daily data from MSCI and match the ratio of market caps to the date on which the individual's data point is drawn. Source: MSCI and Financial Engines.

Relative Returns

International stock returns (MSCI All Countries ex-US returns) in excess of U.S. stock returns (MSCI US) between the period t-1 and t. For each individual, we calculate the cumulative international stock return between t-1 and t, the cumulative return for U.S. stocks between t-1, and t and take the difference. Note that t is defined as the day on which the individual is observed, while t-1 is the previous observation (in annualized percent). Source: MSCI and Financial Engines.

MN Experienced Returns

Following the methodology proposed by Malmendier and Nagel (2011), the experienced returns measure is the weighted average of past returns with weights that depend on an individual's age at time t, how many years ago the return was realized and a parameter that controls for the shape of the weighting function. This paper builds experienced returns based on international stock returns in excess of U.S. stock returns (in annualized percent).

Return Chasing

This variable is constructed using the same methodology as MN Experienced Returns, but uses international stock returns as the relevant past returns.

Flight to Safety

We borrow the flight to safety (FTS) dummy variable for the United States from Baele et al. (2013). They use data on bond and stock returns to measure the occurrence of stress periods in which stock markets 88



decline and liquid benchmark bonds increase in value. Advice Dummy

Dummy variable equal to 1 if the individual has signed the investor service agreement to obtain online advice from Financial Engines. Source: Financial Engines.

Not Stale Advice Dummy

Dummy variable equal to 1 if the individual has accessed the online advice website within the past year. Source: Financial Engines.

Total Equity

Allocation to equities in the overall 401(k) portfolio. The total equity allocation is defined as the combination of investments in Large Cap Stocks, Small and MidCap Stocks, Individual Stocks, Company Stock and International Stocks. This series is individual specific. Source: Financial Engines.

International Equity

Allocation to international equity in the overall 401(k) portfolio. This series is individual specific. Source: Financial Engines.

Zip Code Variables Bachelor's Degree or Higher

Description Bachelor's degree or higher as a percentage of population over 25 years old. Bachelor's degree or higher is the sum of people with a bachelor's degree (hd01_vd22), master's degree (hd01_vd23), professional school degree (hd01_vd24) and doctorate degree (hd01_vd25). This is divided by the total population 25 years and over in the area (hd01_vd01). Census labels are in parentheses. Data is at a Zip code level. Source: U.S. Census Bureau, 2008-2012 American Community Survey - Table B15003: Educational attainment for the population over 25 years and over.

Advanced Degree

Master's degree or higher as a percentage of population over 25 years old. Master's degree or higher is the sum of people with a master's degree (hd01_vd23), professional school degree (hd01_vd24) and doctorate degree (hd01_vd25). This is divided by the total population 25 years and over in the area (hd01_vd01). Census labels are in parentheses. Data is at a Zip code level. Source: U.S. Census Bureau, 2008-2012 American Community Survey - Table B15003 Educational attainment for the population over 25 years and over. 89



Less than college degree

Bachelor's Degree

Less than college degree as a percentage of population over 25 years old. Less than college degree is the sum of people with a regular high school diploma (hd01_vd17), GED high school diploma (hd01_vd18), some college - less than 1 year (hd01_vd19), some college - more than 1 year (hd01_vd20) and associate's degree (hd01_vd21). This sum is divided by the total population 25 years and over in the area (hd01_vd01). Census labels are in parentheses. Data is at a Zip code level. Source: U.S. Census Bureau, 2008-2012 American Community Survey - Table B15003: Educational attainment for the population over 25 years and over. Bachelor's degree as a percentage of population over 25 years old. This variable is defined as people with a bachelor's degree (hd01_vd22) divided by the total population 25 years and over in the area (hd01_vd01). Census labels are in parentheses. Data is at a Zip code level. Source: U.S. Census Bureau, 2008-2012 American Community Survey - Table B15003: Educational attainment for the population over 25 years and over.

Foreign Born Population

Foreign-born population over total population. This variable is defined as Total Foreign Born Population (hd01_vd01) over total population in the area (hc01_vc03). Census labels are in parentheses. Data is at a Zip code level. Source: U.S. Census Bureau, 20072011 American Community Survey - Tables B05007: Place of birth by year of entry by citizenship status for the foreign-born population and DP05: ACS demographic and housing estimates.

Foreign Born Population - Latin America

Foreign-born population from Latin America over total population. This variable is defined as the Latin American born population (hd01_vd28) over total population in the area (hc01_vc03). Census labels are in parentheses. Data is at a Zip code level. Source: U.S. Census Bureau, 2007-2011 American Community Survey - Tables B05007: Place of birth by year of entry by citizenship status for the foreign-born population and DP05: ACS demographic and housing estimates.

Foreign Born Population - Europe

Foreign-born population from Europe over total population. This variable is defined as the European 90



born population (hd01_vd02) over total population in the area (hc01_vc03). Census labels are in parentheses. Data is at a Zip code level. Source: U.S. Census Bureau, 2007-2011 American Community Survey - Tables B05007: Place of birth by year of entry by citizenship status for the foreign-born population and DP05: ACS demographic and housing estimates. Foreign Born Population - Asia

Foreign-born population from Asia over total population. This variable is defined as the Asian born population (hd01_vd15) over total population in the area (hc01_vc03). Census labels are in parentheses. Data is at a Zip code level. Source: U.S. Census Bureau, 2007-2011 American Community Survey - Tables B05007: Place of birth by year of entry by citizenship status for the foreign-born population and DP05: ACS demographic and housing estimates.

Foreign Born Population - Other

Foreign-born population from a region other than Asia, Europe and Latin America over total population. This variable is defined as the “Other” born population (hd01_vd82) over total population in the area (hc01_vc03). Census labels are in parentheses. Data is at a Zip code level. Source: U.S. Census Bureau, 20072011 American Community Survey - Tables B05007: Place of birth by year of entry by citizenship status for the foreign-born population and DP05: ACS demographic and housing estimates.

State Exports/GDP

Export of goods measured as a share of gross domestic product at the state level (ratio is average of 2008-2011 annual data). Source: U.S. Census Bureau and Bureau of Economic Analysis.

State Openness

The sum of exports and imports of goods measured as a share of gross domestic product at the state level (ratio is average of 2008-2011 annual data). Source: U.S. Census Bureau and Bureau of Economic Analysis.

GDP per capita

Per capita real GDP by state (chained 2005 dollars), 2005 to 2011 average. Data is annual. Source: Bureau of Economic Analysis. Real GDP by state (millions of chained 2005 dollars). We take the 2000 to 2005 and 2006 to 2011 growth rates. Data is annual. Source: Bureau of Economic Analysis.

GDP growth

91

Rural

Rural is a categorical variable that takes values 1 to 4 in integer units, with 1 representing the most urban areas and 4 the most isolated. The variable is constructed from the RUCA 2.0 variable in the Zip RUCA Code dataset. More specifically, a Zip code is classified in the following way: (i) urban if RUCA2.0 is 1.0, 1.1, 2.0, 2.1, 3.0, 4.1, 5.1, 7.1, 8.1, or 10.1, (ii) large rural city/town if RUCA2.0 is 4.0, 4.2, 5.0, 5.2, 6.0, or 6.), (iii) small rural town if RUCA2.0 is 7.0, 7.2, 7.3, 7.4, 8.0, 8.2, 8.3, 8.4, 9.0, 9.1, 9.2, and isolated if RUCA2.0 is 10.0, 10.2, 10.3, 10.4, 10.5, or 10.6. Source: RUCA Rural Health Research Center.

Urban

The variable Urban is a dummy variable equal to 1 if RUCA2.0 is equal to 1.0, 1.1, 2.0, 2.1, 3.0, 4.1, 5.1, 7.1, 8.1, or 10.1 (these are the metropolitan areas in the Zip RUCA Code dataset). Data is at the Zip code level. Source: RUCA Rural Health Research Center.

Large Rural

The variable Large Rural is a dummy variable equal to 1 if RUCA2.0 is equal to 4.0, 4.2, 5.0, 5.2, 6.0, or 6.1 (these are the large rural city/town areas in the Zip RUCA Code dataset). Data is at the Zip code level. Source: RUCA Rural Health Research Center.

Small Rural

The variable Small Rural is a dummy variable equal to 1 if RUCA2.0 is equal to 7.0, 7.2, 7.3, 7.4, 8.0, 8.2, 8.3, 8.4, 9.0, 9.1, 9.2 (these are the small rural town areas in the Zip RUCA Code dataset). Data is at the Zip code level. Source: RUCA Rural Health Research Center.

Isolated

The variable Isolated is a dummy variable equal to 1 if RUCA2.0 is equal to 10.0, 10.2, 10.3, 10.4, 10.5, or 10.6 (these are the isolated small rural areas in the Zip RUCA Code dataset). Data is at the Zip code level. Source: RUCA Rural Health Research Center.

Long distance minutes

Number of long distance hours from land lines and mobile phones scaled by total population. Data is at the state level and is the average of the annual data for the 2000-2011 period. Source: FCC.

Distance to International Cities

Distance to international cities is the cumulative distance from each Zip code to London, Tokyo, Toronto and Mexico City (in miles). To calculate the distance from a Zip code to each city, we apply the haversine formula using the latitude and longitude of each point. 92



This formula calculates the great-circle distance between two points (the shortest distance over the earth’s surface), giving an ‘as-the-crow-flies’ distance between the Zip code and the city. We then add the four distances to produce the Zip code's distance to international cities. Source: federalgovernmentZipcodes.us. Financial Literacy

Mean number of correct quiz answers in financial knowledge survey. Multiple choice quiz questions include calculations involving interest rates and inflation, the relationship between bond prices and interest rates, risk and diversification, and the impact of short-term rates on life of a mortgage. Data is at the state level. Source: 2012 National Financial Capability Study Data Tables.

House Value - Zillow (log)

The natural logarithm of house values at the Zip code level deflated to 2005 prices using the Consumer Price Index for All Urban Consumers. We take the average of the deflated monthly data for the period that the Zip code is in the sample (ranges between 2006-2011). Source: Zillow and U.S. Department of Labor: Bureau of Labor Statistics. The natural logarithm of median house values in dollars at the Zip code level. Median house values over USD 1 million are reported as +1,000,000. Since this only affects 158 Zip codes we set them simply to 1,000,000. Source: U.S. Census Bureau, 2008-2012 American Community Survey - Table B25077: Median Housing Value of Owner-Occupied Housing Units (Dollars).

House Value - Census (log)

Firm Variables

Description

Private

Dummy variable that takes the value of 1 if the firm is private and 0 if the firm is public. Source: Capital IQ.

Foreign Headquarter Dummy

Dummy variable that takes the value of 1 if firm's ultimate parent is based in a country outside of the United States. Source: Capital IQ. Dummy variable equal to 1 if firm has a subsidiary in a country outside of the United States. Source: Orbis.

Foreign Subsidiary Dummy % Foreign Subsidiaries

Number of foreign subsidiaries over the total number of subsidiaries in the firm. If company has no subsidiaries, this variable takes the value of zero. Source: Orbis. 93



Industry Openness

The sum of exports and imports of goods measured as a share of gross output by industry (ratio is average of 2000-2011 annual data). Industry is classified at the 3digit NAICS level. Source: U.S. Census Bureau and Bureau of Economic Analysis.

Firm Age (log)

Firm age is calculated as the difference between the current fiscal year and the year the firm was founded. Source: Capital IQ.

Number of Employees (log)

Number of employees in the firm. Use data from Capital IQ only when Compustat data is missing. Given that Compustat reports number of employees in thousands, we multiply the data item “emp” by 1000 in order to be consistent with Capital IQ. We take the average of the annual data for the period that the firm is in the sample (ranges between 2005 and 2011). Source: Compustat and Capital IQ.

Assets (log)

Firm assets in USD million, data item “at” in Compustat, deflated to 2005 prices using the Consumer Price Index for All Urban Consumers. Use data from Capital IQ only when Compustat data is missing. We take the average of the annual data for the period that the firm is in the sample (ranges between 2005 and 2011). Source: Compustat, Capital IQ and U.S. Department of Labor: Bureau of Labor Statistics.

Leverage

Firm total debt over assets, data items (dlc + dltt)/at in Compustat. Use data from Capital IQ only when Compustat data is missing. We take the average of the annual data for the period that the firm is in the sample (ranges between 2005 and 2011). Source: Compustat and Capital IQ.

Sales/Assets

Firm sales over assets, data items “sales” and “at” in Compustat. Use data from Capital IQ only when Compustat data is missing. We take the average of the annual data for the period that the firm is in the sample (ranges between 2005 and 2011). Source: Compustat and Capital IQ.

94

Profitability

Firm net income over assets, data items “ni” and “at” in Compustat. Use data from Capital IQ only when Compustat data is missing. We take the average of the annual data for the period that the firm is in the sample (ranges between 2005 and 2011). Source: Compustat and Capital IQ.

Investment Intensity

Firm capex over assets, data items “capx” and “at” in Compustat. Use data from Capital IQ only when Compustat data is missing. We take the average of the annual data for the period that the firm is in the sample (ranges between 2005 and 2011). Source: Compustat and Capital IQ. Description

Plan Variables Fraction of International Equity Funds

Number of international over Domestic equity funds. The funds are classified as international based on the Lipper categories covering international equity funds, emerging market funds, area or country specific funds. Source: Financial Engines.

Expense Ratio of Intnl-Expense Ratio of Domestic Eq Funds

Difference of the median expense ratio of the international funds and the median expense ratio of the domestic fund offered by the company’s plan(s). Source: Financial Engines. Ratio of the median turnover of the international funds and the median turnover of the domestic fund offered by the company’s plan(s). Source: Financial Engines.

Turnover of Intnl/ Domestic Eq Funds

Alpha of Intnl-Domestic Eq Funds

Fund Age of Intnl/ Domestic Eq Funds

High Fee Plan Dummy

Total Plan Assets

Difference between the median alpha of the international funds and the median alpha of the domestic fund offered by the company’s plan(s). Alphas are calculated relative to a benchmark computed using style analysis with 15 asset classes. Source: Financial Engines. Ratio of the median age of the international funds and the median age of the domestic fund offered by the company’s plan(s). Source: Financial Engines. Dummy variable equal to 1 if the plan scores below the median quality in terms of fees for both the international and the domestic funds, compared to the universe of funds the same categories. Source: Financial Engines and Authors’ calculations. Total asset aggregated across all the plans offered by the firm (USD mn). Source: Financial Engines.

95